FineST inference on demo data

Main goals

  • Goal1: Run FineST on demo dataset that iStar used (as demo).

  • Goal2: Examine the inference performance of FineST at spot level.

[1]:

import os
import time
import warnings
warnings.filterwarnings('ignore')
import numpy as np
import pandas as pd
import torch
import scanpy as sc
import logging
from datetime import datetime
import matplotlib.pyplot as plt
import matplotlib.colors as clr
import json

# import FineST as fst
# from FineST.datasets import dataset
# import FineST.plottings as fstplt
# print("torch version: %s" %torch.__version__)
# print("FineST version: %s" %fst.__version__)

[2]:

## From GPU2
path = '/mnt/lingyu/nfs_share2/Python/'
os.chdir(str(path) + 'FineST/FineST/')
import FineST as fst
from FineST.datasets import dataset
import FineST.plottings as fstplt

[4]:

path = '/mnt/lingyu/nfs_share2/Python/'

fst.setup_seed(666)

if torch.cuda.is_available():
    dev = "cuda:0"
else:
    dev = "cpu"
device = torch.device(dev)

colors = ["#000003",  "#3b0f6f",  "#8c2980",   "#f66e5b", "#fd9f6c", "#fbfcbf"]
cnt_color = clr.LinearSegmentedColormap.from_list('magma', colors, N=256)

0.1 Make logging and save model

[5]:

logging.getLogger().setLevel(logging.INFO)

model_folder = str(path) + 'FineST/FineST_local/Finetune/'
dir_name = model_folder + datetime.now().strftime('%Y%m%d%H%M%S%f')
if not os.path.exists(dir_name):
    os.makedirs(dir_name)
    logger = fst.setup_logger(dir_name)

print(dir_name)

/mnt/lingyu/nfs_share2/Python/FineST/FineST_local/Finetune/20250225230814682723

0.2 Load parameter settings

You can use the parameter_file_path you set or the optimal parameters obtained from your train and test model for your dataset, but here we use our trained parameter_file_path for paper results repeated (saved in FineST/FineST/Parameter/parameters_NPC_P10125.json).

[6]:

# parameter_file_path = str(path) + 'FineST/FineST/parameter/parameters_NPC_P10125.json'
parameter_file_path = str(path) + 'FineST/FineST_local/Parameter/parameters_demo_virchow2.json'
with open(parameter_file_path,"r") as json_file:
    params = json.load(json_file)

logger.info("Load parameters:\n" + json.dumps(params, indent=2))

[2025-02-25 23:08:14] INFO - Load parameters:
{
  "n_encoder_hidden_matrix": 256,
  "n_encoder_hidden_image": 128,
  "n_encoder_latent": 128,
  "n_projection_hidden": 256,
  "n_projection_output": 128,
  "batch_size": 50,
  "batch_size_pair": 64,
  "n_encoder_layers": 2,
  "dropout_rate": 0,
  "training_epoch": 50,
  "inital_learning_rate": 0.25,
  "k_nearest_positives": 0,
  "temperature": 0.03,
  "n_input_image": 1280,
  "w1": 0,
  "w2": 0,
  "w3": 1,
  "w4": 1
}

1. Load ST data and Image embeddings

1.1 Load spatial data

A human Nasopharyngeal carcinoma dataset (NPC_patient_1) from Gong, et al. will be used as a reference dataset, raw_feature_bc_matrix, filtered_feature_bc_matrix, and spatial coordinates (.h5) are publicly available in GSE200310. For easier reuse, we included them in an anndata object which can be loaded directly in FineST Python package.

[7]:

adata_df = pd.read_csv(f"{path}FineST/FineST_local/Dataset/Demo/cnts.tsv", sep='\t', index_col=0)
adata_df

[7]:
FO538757.1 SAMD11 NOC2L KLHL17 PLEKHN1 PERM1 HES4 ISG15 AGRN RNF223 ... F8 FUNDC2 BRCC3 VBP1 CLIC2 SPRY3 VAMP7 USP9Y TMSB4Y NLGN4Y
spot
10x10 0 1 0 0 0 0 0 3 0 0 ... 0 0 0 0 0 0 0 0 0 0
10x11 0 0 0 0 0 0 0 2 0 0 ... 0 0 0 0 0 0 0 0 0 0
10x12 0 0 0 0 0 0 1 1 0 0 ... 0 0 0 1 0 0 0 0 0 0
10x13 0 0 0 0 0 0 1 4 0 0 ... 0 0 0 0 0 0 0 0 0 0
10x14 0 0 0 0 0 0 1 1 0 0 ... 0 0 0 0 0 0 0 0 0 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
9x23 0 0 0 0 0 0 2 1 1 0 ... 0 0 0 0 0 0 0 0 0 0
9x24 0 0 0 0 0 0 1 0 0 0 ... 0 0 1 0 0 0 1 0 0 0
9x25 0 0 0 0 0 0 1 3 1 0 ... 0 1 0 1 0 0 0 0 0 0
9x26 0 0 0 0 0 0 0 0 0 0 ... 0 0 0 0 0 0 0 0 0 0
9x9 0 0 0 0 0 0 0 3 0 0 ... 0 0 0 0 0 0 0 0 0 0

441 rows × 14992 columns

[8]:

loc_df = pd.read_csv(f'{path}FineST/FineST_local/Dataset/Demo/locs-raw.tsv', sep='\t', index_col=0)
loc_df

[8]:
x y
spot
10x10 14160.500000 13648.000000
10x11 14150.500000 12925.500000
10x12 14151.000000 12195.500000
10x13 14126.000000 11465.500000
10x14 14138.500000 10745.500000
... ... ...
9x23 14813.998031 4169.998031
9x24 14801.500000 3440.000000
9x25 14864.000000 2702.500000
9x26 14804.000000 1984.500000
9x9 14821.000000 14343.500000

441 rows × 2 columns

[9]:

# first_col = [1] * len(loc_df)
# second_col = [int(spot.split('x')[0]) for spot in loc_df.index]
# third_col = [int(spot.split('x')[1]) for spot in loc_df.index]

# loc_df_list = pd.DataFrame({
#     0: first_col,
#     1: second_col,
#     2: third_col,
#     3: loc_df['y'].values,
#     4: loc_df['x'].values
# }, index=loc_df.index)
# print(loc_df_list.head())
# loc_df_list.to_csv(f'{path}FineST/FineST_local/Dataset/Demo/tissue_positions_list.csv', header=False)

[10]:

# sacle_factor = 0.14
# loc_df_sacle = loc_df*sacle_factor
# loc_df_sacle.columns = ['pxl_row_in_fullres', 'pxl_col_in_fullres']
# print(loc_df_sacle.head())
# loc_df_sacle.to_csv(f'{path}FineST/FineST_local/Dataset/Demo/locs_raw.csv')

[11]:

print(loc_df['x'].max(), loc_df['y'].max())

20057.5 15798.0

[10]:

from PIL import Image
Image.MAX_IMAGE_PIXELS = None

img = Image.open(f'{path}FineST/FineST_local/Dataset/Demo/he-raw.jpg')
img = np.array(img)

# Display the image
imgplot = plt.imshow(img)
plt.show()
_images/Demo_Train_Impute_count_18_0.png 
[12]:

from scipy.sparse import csr_matrix
adata = sc.AnnData(np.zeros((adata_df.shape[0], adata_df.shape[1])))
adata.X = csr_matrix(adata_df, dtype=np.float32)
adata.obs_names = adata_df.index
adata.var_names = adata_df.columns
adata.obsm['spatial'] = np.array(loc_df.loc[:, ['x', 'y']])
adata.obs['in_tissue'] = pd.DataFrame([1]*adata_df.shape[0], index=adata_df.index, columns=['in_tissue'])
adata

[12]:

AnnData object with n_obs × n_vars = 441 × 14992
    obs: 'in_tissue'
    obsm: 'spatial'

[13]:

print(adata.obsm['spatial'][:5])
print(adata.obsm['spatial'][:,0].max(), adata.obsm['spatial'][:,1].max())

[[14160.5 13648. ]
 [14150.5 12925.5]
 [14151.  12195.5]
 [14126.  11465.5]
 [14138.5 10745.5]]
20057.5 15798.0

[14]:

print(13648.0-12925.5)
print(12925.5-12195.5)
print(12195.5-11465.5)

722.5
730.0
730.0

[15]:

patch_size_x = fst.patch_size(adata, p=6, dir='x')
patch_size_y = fst.patch_size(adata, p=6, dir='y')
print(patch_size_x)
print(patch_size_y)

94      707.500000
115     745.000000
156    1387.500000
176     747.500000
197     725.001969
Name: 0, dtype: float64
311     747.0
304    4375.0
310    4340.0
309     690.0
308     737.5
Name: 1, dtype: float64

[16]:

# adata_example = dataset.NPC()
# print(adata_example)

1.2 Selected LR genes

In FineST, we only considers 963 human genes involved in ligand-receptor pairs from CellChatDB (version.1.1.3) (mouse: 2,022 pairs, human: 1,940 pairs, zebrafish: 2,774 pairs), and extract the overlap gene’s expression profile.

[17]:

LR_gene_path = str(path)+'FineST/FineST/Dataset/LRgene/LRgene_CellChatDB_baseline.csv'
adata = fst.adata_LR(adata, LR_gene_path)
print(adata)

AnnData object with n_obs × n_vars = 441 × 647
    obs: 'in_tissue'
    obsm: 'spatial'

1.3 Data preprocess

After making an intersection with the LR genes and filtering out genes that are detected in less than 10 cells, 596 genes from the NPC datasetare used for training and prediction.

[18]:

adata_count = adata.copy()
adata = fst.adata_preprocess(adata_count, keep_raw=True, normalize=False, min_cells=0)
print(adata)

gene_hv = np.array(adata.var_names)
matrix = fst.adata2matrix(adata, gene_hv)
matrix

AnnData object with n_obs × n_vars = 441 × 647
    obs: 'in_tissue'
    var: 'mt', 'n_cells'
    obsm: 'spatial'
(441, 647)

[18]:
TGFB1 TGFBR1 TGFBR2 TGFB2 TGFB3 ACVR1B ACVR1C ACVR1 BMP2 BMPR1A ... SEMA6D KDR TREM2 SEMA6A SEMA6B SEMA7A PLXNC1 SIGLEC1 THY1 VCAM1
10x10 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0
10x11 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
10x12 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 2.0 0.0 0.0 0.0 4.0 0.0
10x13 2.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0
10x14 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
9x23 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3.0 0.0
9x24 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0
9x25 1.0 1.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0
9x26 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
9x9 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0

441 rows × 647 columns

[19]:

contains_nan = matrix.isnull().values.any()
print(f"Does adata.X contain NaN values? {contains_nan}")

Does adata.X contain NaN values? False

1.4 Order by image file name

Load image embedding from Setp0 HIPT_image_feature_extract.py. For NPC dataset, the image embedding file can be downloaded Goole Drive (within spots: AH_Patient1_pth_64_16, between spots: NEW_AH_Patient1_pth_64_16, single-nuclei: sc_Patient1_pth_16_16).

[20]:

file_paths = sorted(os.listdir(str(path) + 'FineST/FineST_local/Dataset/Demo/HIPT/Demo_pth_56_14/'))
print("Image embedding file (first three .pth):\n", file_paths[:3])

Image embedding file (first three .pth):
 ['Demo_FineST_10415.5_7065.499999999999.pth', 'Demo_FineST_10418.0_5607.5.pth', 'Demo_FineST_10418.499999999998_2705.0000000000005.pth']

[21]:

## Image patch position
position_image = fst.get_image_coord(file_paths, dataset_class="VisiumHD")

## convert numeric and save %2.f
position_image = position_image.apply(pd.to_numeric)
position_image = position_image.round(1)

print("The coords of image patch: \n",position_image.shape)

contains_nan = position_image.isnull().values.any()
print(f"Does adata.X contain NaN values? {contains_nan}")

print(position_image.head())

position_image

The coords of image patch:
 (441, 2)
Does adata.X contain NaN values? False
   pixel_x  pixel_y
0   7065.5  10415.5
1   5607.5  10418.0
2   2705.0  10418.5
3   1242.5  10421.0
4   6352.5  10425.5

[21]:
pixel_x pixel_y
0 7065.5 10415.5
1 5607.5 10418.0
2 2705.0 10418.5
3 1242.5 10421.0
4 6352.5 10425.5
... ... ...
436 6370.5 9770.5
437 15798.0 9772.5
438 10750.5 9775.5
439 13635.5 9775.5
440 15093.0 9787.5

441 rows × 2 columns

[22]:

## ST spot position
position = pd.read_csv(str(path)+'FineST/FineST_local/Dataset/Demo/tissue_positions_list.csv', header=None)
position = position.rename(columns={position.columns[-2]: 'pixel_x',
                                    position.columns[-1]: 'pixel_y'})
print("The coords of ST spot: \n", position.shape)

contains_nan = position.isnull().values.any()
print(f"Does adata.X contain NaN values? {contains_nan}")

## convert numeric and save %2.f
position[['pixel_x', 'pixel_y']] = position[['pixel_x', 'pixel_y']].apply(pd.to_numeric)
position[['pixel_x', 'pixel_y']] = position[['pixel_x', 'pixel_y']].round(1)

print(position.head())
position

The coords of ST spot:
 (441, 6)
Does adata.X contain NaN values? False
       0  1   2   3  pixel_x  pixel_y
0  10x10  1  10  10  13648.0  14160.5
1  10x11  1  10  11  12925.5  14150.5
2  10x12  1  10  12  12195.5  14151.0
3  10x13  1  10  13  11465.5  14126.0
4  10x14  1  10  14  10745.5  14138.5

[22]:
0 1 2 3 pixel_x pixel_y
0 10x10 1 10 10 13648.0 14160.5
1 10x11 1 10 11 12925.5 14150.5
2 10x12 1 10 12 12195.5 14151.0
3 10x13 1 10 13 11465.5 14126.0
4 10x14 1 10 14 10745.5 14138.5
... ... ... ... ... ... ...
436 9x23 1 9 23 4170.0 14814.0
437 9x24 1 9 24 3440.0 14801.5
438 9x25 1 9 25 2702.5 14864.0
439 9x26 1 9 26 1984.5 14804.0
440 9x9 1 9 9 14343.5 14821.0

441 rows × 6 columns

[23]:

position_image.head()
print(position_image['pixel_x'].max(), position_image['pixel_y'].max())

15798.0 20057.5

[24]:

position.head()
print(position['pixel_x'].max(), position['pixel_y'].max())

15798.0 20057.5

[25]:

## merge position
position_image = fst.image_coord_merge(position_image, position, dataset_class = 'Visium')

contains_nan = position_image.isnull().values.any()
print(f"Does adata.X contain NaN values? {contains_nan}")

position_image

Does adata.X contain NaN values? False

[25]:
0 1 x y pixel_x pixel_y
0 15x19 1 15 19 7065.5 10415.5
1 15x21 1 15 21 5607.5 10418.0
2 15x25 1 15 25 2705.0 10418.5
3 15x27 1 15 27 1242.5 10421.0
4 15x20 1 15 20 6352.5 10425.5
... ... ... ... ... ... ...
436 16x20 1 16 20 6370.5 9770.5
437 16x7 1 16 7 15798.0 9772.5
438 16x14 1 16 14 10750.5 9775.5
439 16x10 1 16 10 13635.5 9775.5
440 16x8 1 16 8 15093.0 9787.5

441 rows × 6 columns

[26]:

position_order = fst.update_st_coord(position_image)
print("The coords of image patch (merged): \n", position_order.shape)
print(position_order.head())

The coords of image patch (merged):
 (441, 4)
   pixel_y  pixel_x  array_row  array_col
0  10415.5   7065.5         19         15
1  10418.0   5607.5         21         15
2  10418.5   2705.0         25         15
3  10421.0   1242.5         27         15
4  10425.5   6352.5         20         15

[27]:

# ## merge position
# position_image = fst.image_coord_merge(position_image, position, dataset_class = 'Visium')
# position_order = fst.update_st_coord(position_image)
# print("The coords of image patch (merged): \n", position_order.shape)
# print(position_order.head())

[28]:

print(position_order['pixel_y'].max(), position_order['pixel_x'].max())

20057.5 15798.0

[29]:

# print(position_order['pixel_y'].max(), position_order['pixel_x'].max())

[30]:

## save the position data  -- train model used
position_order.to_csv(str(path)+'FineST/FineST_local/Dataset/Demo/ContrastDemoLRgene/position_order.csv', index=False, header=False)

1.5 Order matrix row/barcode, spatial coordinates by image coordinates

[31]:

spotID_order = np.array(position_image[0])
matrix_order, matrix_order_df = fst.sort_matrix(matrix, position_image, spotID_order, gene_hv)
matrix_order_df

[31]:
TGFB1 TGFBR1 TGFBR2 TGFB2 TGFB3 ACVR1B ACVR1C ACVR1 BMP2 BMPR1A ... SEMA6D KDR TREM2 SEMA6A SEMA6B SEMA7A PLXNC1 SIGLEC1 THY1 VCAM1
15x19 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
15x21 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
15x25 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0
15x27 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 5.0 0.0
15x20 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
16x20 1.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
16x7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
16x14 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
16x10 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
16x8 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0

441 rows × 647 columns

[32]:

contains_nan = matrix.isnull().values.any()
print(f"Does adata.X contain NaN values? {contains_nan}")

contains_nan = matrix_order_df.isnull().values.any()
print(f"Does adata.X contain NaN values? {contains_nan}")

Does adata.X contain NaN values? False
Does adata.X contain NaN values? False

[33]:

## save gene expression mateix -- train model used
np.save(str(path)+'FineST/FineST_local/Dataset/Demo/ContrastDemoLRgene/harmony_matrix.npy', matrix_order_df.T)

1.6 Update adata with new barcodes and lications

[34]:

## adata - count
adata = fst.update_adata_coord(adata, matrix_order_df, position_image)
print(adata)
print(adata.obsm['spatial'][:5])

AnnData object with n_obs × n_vars = 441 × 647
    obs: 'in_tissue', 'array_row', 'array_col'
    var: 'mt', 'n_cells'
    obsm: 'spatial'
[[10415.5  7065.5]
 [10418.   5607.5]
 [10418.5  2705. ]
 [10421.   1242.5]
 [10425.5  6352.5]]

[35]:

## adata - normalized
adata_count = adata.copy()
adata_norma = fst.adata_preprocess(adata_count, keep_raw=True, normalize=True, min_cells=0)
print(adata_norma)
adata_norma.to_df().head()

AnnData object with n_obs × n_vars = 441 × 647
    obs: 'in_tissue', 'array_row', 'array_col'
    var: 'mt', 'n_cells'
    uns: 'log1p'
    obsm: 'spatial'

[35]:
TGFB1 TGFBR1 TGFBR2 TGFB2 TGFB3 ACVR1B ACVR1C ACVR1 BMP2 BMPR1A ... SEMA6D KDR TREM2 SEMA6A SEMA6B SEMA7A PLXNC1 SIGLEC1 THY1 VCAM1
15x19 0.0 0.000000 0.0 0.57145 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.00000 0.0 0.0 0.000000 0.0 0.0 0.0 0.571450 0.0
15x21 0.0 0.639304 0.0 0.00000 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.00000 0.0 0.0 0.000000 0.0 0.0 0.0 0.000000 0.0
15x25 0.0 0.507248 0.0 0.00000 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.00000 0.0 0.0 0.000000 0.0 0.0 0.0 0.842183 0.0
15x27 0.0 0.000000 0.0 0.00000 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.00000 0.0 0.0 0.706941 0.0 0.0 0.0 1.814644 0.0
15x20 0.0 0.000000 0.0 0.00000 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.37753 0.0 0.0 0.000000 0.0 0.0 0.0 1.041940 0.0

5 rows × 647 columns

[36]:

adata_count.to_df()

[36]:
TGFB1 TGFBR1 TGFBR2 TGFB2 TGFB3 ACVR1B ACVR1C ACVR1 BMP2 BMPR1A ... SEMA6D KDR TREM2 SEMA6A SEMA6B SEMA7A PLXNC1 SIGLEC1 THY1 VCAM1
15x19 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
15x21 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
15x25 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0
15x27 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 5.0 0.0
15x20 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 4.0 0.0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
16x20 1.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
16x7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
16x14 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
16x10 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
16x8 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0

441 rows × 647 columns

[37]:

## save the original adata
patientxy = 'Demo_FineST'
adata_count.write_h5ad(str(path)+'FineST/FineST_local/Dataset/ImputData/'+str(patientxy)+'/'+str(patientxy)+'_adata_count.h5ad')
adata_norma.write_h5ad(str(path)+'FineST/FineST_local/Dataset/ImputData/'+str(patientxy)+'/'+str(patientxy)+'_adata_norma.h5ad')

Note The above cells for section 1.5 only need to be run once. They aim to generate the ordered gene expression profile harmony_matrix.npy and the ordered ST spot coordinates position_order.csv according to image pixel coordinates, and save the normalized gene expression data patient1_adata_orignal.

[39]:

os.chdir(f'{path}FineST/FineST_local/Dataset/ImputData/Demo_FineST/Figure')
!pwd

/mnt/lingyu/nfs_share2/Python/FineST/FineST_local/Dataset/ImputData/Demo_FineST/Figure

[65]:

fstplt.gene_expr(adata_count, adata_count.to_df(), gene_selet='ERBB2', marker='o',
                 s=80, figsize=(5, 4), cnt_color='turbo', save_path='ERBB2_expr_count.pdf')
fstplt.gene_expr(adata_norma, adata_norma.to_df(), gene_selet='ERBB2', marker='o',
                 s=80, figsize=(5, 4), cnt_color='turbo', save_path='ERBB2_expr_norml.pdf')
_images/Demo_Train_Impute_count_55_0.png
_images/Demo_Train_Impute_count_55_1.png

2 Train and Test model

2.1 Data loader and splitting: 80% for train and 20% for test

Input: three input file pathway:
- image_embed_path: image embedding from Setp0 HIPT_image_feature_extract.py
- spatial_pos_path: ordered ST spot coords, in ./Dataset/Demo/ContrastDemoLRgene/position_order.csv
- reduced_mtx_path: ordered gene expression, in ./Dataset/Demo/ContrastDemoLRgene/harmony_matrix.np

For NPC dataset, the image embedding file can be downloaded Goole Drive.

[41]:

image_embed_path_NPC = str(path)+'FineST/FineST_local/Dataset/Demo/HIPT/Demo_pth_56_14/*.pth'
spatial_pos_path = str(path)+'FineST/FineST_local/Dataset/Demo/ContrastDemoLRgene/position_order.csv'
reduced_mtx_path = str(path)+'FineST/FineST_local/Dataset/Demo/ContrastDemoLRgene/harmony_matrix.npy'

2.2 Train and test model on within spot

[41]:

# train_loader, test_loader = fst.build_loaders(batch_size=params['batch_size'],
#                                               image_embed_path=image_embed_path_NPC,
#                                               spatial_pos_path=spatial_pos_path,
#                                               reduced_mtx_path=reduced_mtx_path,
#                                               image_clacss='Virchow2',
#                                               dataset_class='Visium16')

[42]:

# input_spot_test, input_image_test, input_coord_test, input_row_test, input_col_test = fst.extract_test_data(train_loader)
# print("input_spot_test:", input_spot_test.shape)
# print("input_image_test:", input_image_test.shape)
# print(input_image_test)
# print(input_spot_test)

[43]:

# input_spot_test, input_image_test, input_coord_test, input_row_test, input_col_test = fst.extract_test_data(test_loader)
# print("input_spot_test:", input_spot_test.shape)
# print("input_image_test:", input_image_test.shape)
# print(input_image_test)
# print(input_spot_test)

[44]:

# all_dataset = fst.build_loaders_inference(batch_size=adata.shape[0],
#                                           image_embed_path=image_embed_path_NPC,
#                                           spatial_pos_path=spatial_pos_path,
#                                           reduced_mtx_path=reduced_mtx_path,
#                                           image_clacss='Virchow2',
#                                           dataset_class='Visium16')

[43]:

## add params
params['n_input_matrix'] = len(gene_hv)

## init the model
model = fst.FineSTModel(n_input_matrix=params['n_input_matrix'],
                              n_input_image=params['n_input_image'],
                              n_encoder_hidden_matrix=params["n_encoder_hidden_matrix"],
                              n_encoder_hidden_image=params["n_encoder_hidden_image"],
                              n_encoder_latent=params["n_encoder_latent"],
                              n_projection_hidden=params["n_projection_hidden"],
                              n_projection_output=params["n_projection_output"],
                              n_encoder_layers=params["n_encoder_layers"]).to(device)
## Load the data
train_loader, test_loader = fst.build_loaders(batch_size=params['batch_size'],
                                              image_embed_path=image_embed_path_NPC,
                                              spatial_pos_path=spatial_pos_path,
                                              reduced_mtx_path=reduced_mtx_path,
                                              image_clacss='Virchow2',
                                              dataset_class='Visium16')
## Set optimizer
optimizer = torch.optim.SGD(model.parameters(), lr=params['inital_learning_rate'],
                            momentum=0.9, weight_decay=5e-4)
## Load loss function
l = fst.ContrastiveLoss(temperature=params['temperature'],
                        w1=params['w1'], w2=params['w2'], w3=params['w3'], w4=params['w4'])

## train model
(dir_name,
 train_losses, test_losses,
 best_epoch, best_loss) = fst.train_model_fst(params, model, train_loader, test_loader,
                                          optimizer, l, dir_name, logger, dataset_class='Visium16')

Finished loading all files
train/test split completed
352 89
***** Finished building loaders *****

[2025-02-25 21:34:50] INFO - Begin Training ...
[2025-02-25 21:34:50] INFO - epoch [1/50]

Epoch: 1
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 121.28it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............
--- 2.0237627029418945 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 150.50it/s]
[2025-02-25 21:34:52] INFO - Saved Best epoch & Best Model! Loss: [0: 2.2767348289489746]
[2025-02-25 21:34:52] INFO - epoch [2/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [0: 2.2767348289489746]
Epoch: 2
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 122.83it/s]

image_tensor: torch.Size([350, 16, 1280])



checkNeighbors.............
--- 0.7048609256744385 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 107.75it/s]
[2025-02-25 21:34:53] INFO - Saved Best epoch & Best Model! Loss: [1: 2.208057403564453]
[2025-02-25 21:34:53] INFO - epoch [3/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [1: 2.208057403564453]
Epoch: 3
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 160.56it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7539036273956299 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 109.02it/s]
[2025-02-25 21:34:53] INFO - Saved Best epoch & Best Model! Loss: [2: 2.1968281269073486]
[2025-02-25 21:34:53] INFO - epoch [4/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [2: 2.1968281269073486]
Epoch: 4
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 175.22it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7813408374786377 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 124.52it/s]
[2025-02-25 21:34:54] INFO - Saved Best epoch & Best Model! Loss: [3: 2.1609601974487305]
[2025-02-25 21:34:54] INFO - epoch [5/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [3: 2.1609601974487305]
Epoch: 5
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 143.26it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7513885498046875 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 116.65it/s]
[2025-02-25 21:34:55] INFO - Saved Best epoch & Best Model! Loss: [4: 2.12542724609375]
[2025-02-25 21:34:55] INFO - epoch [6/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [4: 2.12542724609375]
Epoch: 6
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 146.56it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8449461460113525 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 123.50it/s]
[2025-02-25 21:34:56] INFO - Saved Best epoch & Best Model! Loss: [5: 2.1060452461242676]
[2025-02-25 21:34:56] INFO - epoch [7/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [5: 2.1060452461242676]
Epoch: 7
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 165.41it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.740849494934082 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 111.74it/s]
[2025-02-25 21:34:57] INFO - Saved Best epoch & Best Model! Loss: [6: 2.0971837043762207]
[2025-02-25 21:34:57] INFO - epoch [8/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [6: 2.0971837043762207]
Epoch: 8
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 148.75it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7784814834594727 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 111.65it/s]
[2025-02-25 21:34:58] INFO - Saved Best epoch & Best Model! Loss: [7: 2.0714845657348633]
[2025-02-25 21:34:58] INFO - epoch [9/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [7: 2.0714845657348633]
Epoch: 9
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 152.04it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8362410068511963 seconds ---
test model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 77.58it/s]
[2025-02-25 21:34:59] INFO - epoch [10/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 10
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 155.53it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8078935146331787 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 113.65it/s]
[2025-02-25 21:35:00] INFO - Saved Best epoch & Best Model! Loss: [9: 2.0481417179107666]
[2025-02-25 21:35:00] INFO - epoch [11/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [9: 2.0481417179107666]
Epoch: 11
train model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 88.79it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............
--- 0.7842123508453369 seconds ---
test model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 81.59it/s]
[2025-02-25 21:35:01] INFO - epoch [12/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 12
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 165.77it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8046705722808838 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 107.02it/s]
[2025-02-25 21:35:02] INFO - Saved Best epoch & Best Model! Loss: [11: 2.001929759979248]
[2025-02-25 21:35:02] INFO - epoch [13/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [11: 2.001929759979248]
Epoch: 13
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 159.13it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7186684608459473 seconds ---
test model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 84.76it/s]
[2025-02-25 21:35:02] INFO - epoch [14/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 14
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 157.67it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6941437721252441 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 218.37it/s]
[2025-02-25 21:35:03] INFO - epoch [15/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 15
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 157.40it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7442679405212402 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 106.56it/s]
[2025-02-25 21:35:04] INFO - epoch [16/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 16
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 136.02it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7217855453491211 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 134.47it/s]
[2025-02-25 21:35:05] INFO - epoch [17/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 17
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 163.53it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.712899923324585 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 132.60it/s]
[2025-02-25 21:35:06] INFO - epoch [18/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 18
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 161.60it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7112278938293457 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 118.34it/s]
[2025-02-25 21:35:06] INFO - epoch [19/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 19
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 164.77it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.775050163269043 seconds ---
test model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 75.45it/s]
[2025-02-25 21:35:07] INFO - epoch [20/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 20
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 153.31it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7141134738922119 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 118.20it/s]
[2025-02-25 21:35:08] INFO - Saved Best epoch & Best Model! Loss: [19: 1.9647986888885498]
[2025-02-25 21:35:08] INFO - epoch [21/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [19: 1.9647986888885498]
Epoch: 21
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 139.38it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.827967643737793 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 101.69it/s]
[2025-02-25 21:35:09] INFO - epoch [22/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 22
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 102.93it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.829883337020874 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 128.02it/s]
[2025-02-25 21:35:10] INFO - epoch [23/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 23
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 120.65it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6942675113677979 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 178.20it/s]
[2025-02-25 21:35:11] INFO - epoch [24/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 24
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 104.88it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6961431503295898 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 114.32it/s]
[2025-02-25 21:35:12] INFO - epoch [25/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 25
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 153.51it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7497637271881104 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 114.34it/s]
[2025-02-25 21:35:12] INFO - epoch [26/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 26
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 156.59it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7248537540435791 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 144.09it/s]
[2025-02-25 21:35:13] INFO - epoch [27/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 27
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 128.72it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.688248872756958 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 113.26it/s]
[2025-02-25 21:35:14] INFO - epoch [28/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 28
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 160.63it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8070394992828369 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 111.73it/s]
[2025-02-25 21:35:15] INFO - epoch [29/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 29
train model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 90.35it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7255241870880127 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 114.13it/s]
[2025-02-25 21:35:16] INFO - epoch [30/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 30
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 161.64it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7546327114105225 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 106.43it/s]
[2025-02-25 21:35:17] INFO - epoch [31/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 31
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 160.81it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7888367176055908 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 129.68it/s]
[2025-02-25 21:35:17] INFO - epoch [32/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 32
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 164.43it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.74188232421875 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 115.24it/s]
[2025-02-25 21:35:18] INFO - epoch [33/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 33
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 161.98it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.74884033203125 seconds ---
test model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 85.96it/s]
[2025-02-25 21:35:19] INFO - epoch [34/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 34
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 160.81it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.825676679611206 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 110.73it/s]
[2025-02-25 21:35:20] INFO - epoch [35/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 35
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 161.75it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8052530288696289 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 162.53it/s]
[2025-02-25 21:35:21] INFO - epoch [36/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 36
train model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 98.93it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8674955368041992 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 115.46it/s]
[2025-02-25 21:35:22] INFO - epoch [37/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 37
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 171.20it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8214435577392578 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 133.93it/s]
[2025-02-25 21:35:23] INFO - epoch [38/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 38
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 152.64it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6625683307647705 seconds ---
test model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 54.26it/s]
[2025-02-25 21:35:23] INFO - epoch [39/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 39
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 169.37it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7014079093933105 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 105.03it/s]
[2025-02-25 21:35:24] INFO - epoch [40/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 40
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 167.55it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6767313480377197 seconds ---
test model

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 83.49it/s]
[2025-02-25 21:35:25] INFO - epoch [41/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 41
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 120.50it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7199041843414307 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 122.39it/s]
[2025-02-25 21:35:26] INFO - epoch [42/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 42
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 171.88it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6600198745727539 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 128.52it/s]
[2025-02-25 21:35:27] INFO - Saved Best epoch & Best Model! Loss: [41: 1.9583055973052979]
[2025-02-25 21:35:27] INFO - epoch [43/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Saved Best epoch & Best Model! Loss: [41: 1.9583055973052979]
Epoch: 43
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 154.45it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6814134120941162 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 125.22it/s]
[2025-02-25 21:35:27] INFO - epoch [44/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 44
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 101.41it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............
--- 0.7222883701324463 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 146.91it/s]
[2025-02-25 21:35:28] INFO - epoch [45/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 45
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 108.17it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7809743881225586 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 127.61it/s]
[2025-02-25 21:35:29] INFO - epoch [46/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 46
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 153.50it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6611125469207764 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 120.95it/s]
[2025-02-25 21:35:30] INFO - epoch [47/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 47
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 141.68it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.8554863929748535 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 113.82it/s]
[2025-02-25 21:35:31] INFO - epoch [48/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 48
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 160.35it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.6332738399505615 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 137.83it/s]
[2025-02-25 21:35:32] INFO - epoch [49/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 49
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 163.32it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7280764579772949 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 107.39it/s]
[2025-02-25 21:35:32] INFO - epoch [50/50]

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Epoch: 50
train model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:00<00:00, 162.78it/s]

image_tensor: torch.Size([350, 16, 1280])
checkNeighbors.............



--- 0.7223777770996094 seconds ---
test model

100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 114.01it/s]
[2025-02-25 21:35:33] INFO - Done!, Best epoch & Best Model! Loss: [41: 1.9583055973052979]
[2025-02-25 21:35:33] INFO - Finished Training

image_tensor: torch.Size([50, 16, 1280])
checkNeighbors.............
Done!, final loss: 1.9583055973052979
Best epoch: 41
--- 43.54628777503967 seconds ---

[44]:

pwd

[44]:

'/mnt/lingyu/nfs_share2/Python/FineST/FineST_local/Dataset/ImputData/Demo_FineST/Figure'

[45]:

fst.loss_curve(train_losses, test_losses, best_epoch, best_loss, max_step=5, min_step=1,
               fig_size=(5, 4), format='svg', save_path='Loss_cvrve_demo_allLRgs.svg')
_images/Demo_Train_Impute_count_67_0.png

Note The above cell for section 2.2 needs to be run more than once by setting different parameters in the given .json file, if the correlation in section 3.4 is not satisfying. If one obtains a good performance, one can save the dir_name and then directly use it to load the trained model and infer gene expression, and doesn’t need to run section 2.2 again. Recommended parameter fine-tuning and its range: training_epoch: 40~60; temperature: 0.01-0.05.

3. Inference, Imputation and Evaluation on within spot

3.1 Infer the gene expression of within spots

[46]:

dir_name

[46]:

'/mnt/lingyu/nfs_share2/Python/FineST/FineST_local/Finetune/20250225213442149378'

You can use the dir_name just obtained from the above cell, but here we use our trained dir_name for paper results repeated (saved in FineST/FineST/Finetune/20250225212431608560).

[47]:

# dir_name = str(path)+'FineST/FineST_local/Finetune/2025025212431608560' # min_cells=1
dir_name = str(path)+'FineST/FineST_local/Finetune/20250225213442149378' # min_cells=0

[48]:

## Load the trained model
model = fst.load_model(dir_name, parameter_file_path, params, gene_hv)

## Load test data
test_loader = fst.build_loaders_inference(batch_size=adata.shape[0],
                                          image_embed_path=image_embed_path_NPC,
                                          spatial_pos_path=spatial_pos_path,
                                          reduced_mtx_path=reduced_mtx_path,
                                          image_clacss='Virchow2',
                                          dataset_class='Visium16')
## Inference
(matrix_profile,
 reconstructed_matrix,
 recon_ref_adata_image_f2,
 reconstructed_matrix_reshaped,
 input_coord_all) = fst.infer_model_fst(model, test_loader, logger, dataset_class='Visium16')

[2025-02-25 23:09:16] INFO - Running inference task...

Finished loading all files
***** Finished building loaders_inference *****
device cuda:0
***** Begin perform_inference: ******

100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 17.52it/s]

***** batch_size=adata.shape[0] doen't effect *****
torch.Size([441, 647])
torch.Size([441, 16, 1280])
1
1
1
***** *****
Finished extractting test data


[2025-02-25 23:09:17] INFO - Running inference task DONE!

--- 0.35297513008117676 seconds for inference within spots ---
Reconstructed_matrix_reshaped shape:  torch.Size([7056, 647])

[49]:

print(matrix_profile)
print(reconstructed_matrix)
print(recon_ref_adata_image_f2)
print(input_coord_all)

[[0. 0. 0. ... 0. 1. 0.]
 [0. 2. 0. ... 0. 0. 0.]
 [0. 2. 0. ... 0. 4. 0.]
 ...
 [0. 0. 0. ... 0. 0. 0.]
 [0. 0. 0. ... 0. 1. 0.]
 [1. 0. 0. ... 0. 1. 0.]]
[[0.09811556 0.00925002 0.00976041 ... 0.00925135 0.05734217 0.00821138]
 [0.00968614 0.00853146 0.00715685 ... 0.00801609 0.00916301 0.00630441]
 [0.09017586 0.16494617 0.01208244 ... 0.00820829 0.57795185 0.0060407 ]
 ...
 [0.05360792 0.00895031 0.00825155 ... 0.00870989 0.00960729 0.00855205]
 [0.00834366 0.00856199 0.0083601  ... 0.0051428  0.04961061 0.00600153]
 [0.04252668 0.00902818 0.00875428 ... 0.00944272 0.00998054 0.00824875]]
[[0.08125865 0.01641767 0.01061217 ... 0.00793805 0.02972159 0.00814416]
 [0.01345231 0.02695793 0.00759121 ... 0.00840887 0.01990831 0.00668759]
 [0.03220735 0.1597349  0.03828571 ... 0.00817165 0.21686108 0.00784339]
 ...
 [0.0199759  0.00892561 0.00805292 ... 0.00821835 0.00862657 0.00812187]
 [0.01728782 0.02685513 0.03752211 ... 0.00704592 0.01613463 0.00650321]
 [0.04118104 0.00872994 0.0084417  ... 0.00880175 0.05676795 0.00864514]]
[[tensor([10415.5000, 10418.0000, 10418.5000, 10421.0000, 10425.5000, 10428.0000,
        10428.0000,  1043.5000, 10432.5000, 10433.0000, 10433.5000, 10435.5000,
        10435.5000, 10438.5000, 10438.5000, 10440.5000, 10443.0000, 10445.5000,
        10445.5000, 10450.5000,  1046.0000, 10563.5000, 11153.0000, 11158.5000,
        11158.5000, 11163.5000, 11166.0000, 11166.0000, 11170.5000, 11170.5000,
        11170.5000, 11170.5000, 11173.0000, 11173.0000, 11173.5000, 11173.5000,
        11175.5000, 11175.5000, 11180.5000, 11183.0000, 11185.5000, 11188.0000,
        11188.0000, 11855.5000, 11866.0000, 11870.5000, 11871.0000, 11871.0000,
        11871.0000, 11873.0000, 11873.4000, 11873.5000, 11873.5000, 11876.0000,
        11878.0000, 11878.0000, 11878.0000, 11878.5000, 11878.5000, 11880.5000,
        11888.0000, 11890.5000, 11890.5000, 12508.0000, 12558.5000, 12558.5000,
        12558.5000, 12563.0000, 12563.0000, 12566.0000, 12568.5000, 12570.5000,
        12573.0000, 12573.0000, 12573.5000, 12576.0000, 12578.0000, 12578.5000,
        12578.5000, 12580.5000, 12583.5000, 12588.5000, 13373.5000, 13378.5000,
        13381.0000, 13383.0000, 13388.5000, 13388.5000, 13388.5000, 13391.0000,
        13395.5000, 13398.5000, 13401.0000, 13401.0000, 13406.0000, 13408.0000,
        13408.5000, 13410.5000, 13413.0000, 13413.5000, 14101.0000, 14126.0000,
        14129.0000, 14133.5000, 14136.0000, 14136.0000, 14138.5000, 14138.5000,
        14138.5000, 14138.5000, 14138.5000, 14143.0000, 14145.5000, 14146.0000,
        14149.0000, 14150.5000, 14151.0000, 14151.0000, 14156.0000, 14160.5000,
        14798.5000, 14798.5000, 14801.0000, 14801.5000, 14803.5000, 14804.0000,
        14806.0000, 14808.5000, 14811.0000, 14811.0000, 14813.5000, 14813.5000,
        14814.0000, 14816.0000, 14818.5000, 14821.0000, 14821.5000, 14864.0000,
        15528.5000, 15563.5000, 15564.0000, 15564.0000, 15569.0000, 15569.0000,
        15573.5000, 15573.5000, 15575.9000, 15576.0000, 15576.5000, 15579.0000,
        15581.5000, 15581.5000, 15581.5000, 15583.5000, 15588.5000, 15601.0000,
        16286.5000, 16289.0000, 16298.5000, 16299.0000, 16299.0000, 16301.5000,
        16301.5000, 16306.5000, 16306.5000, 16306.5000, 16306.5000, 16309.0000,
        16309.0000, 16311.0000, 16313.5000, 16321.0000, 16326.0000, 16889.0000,
        17034.0000, 17041.5000, 17044.0000, 17046.5000, 17046.5000, 17049.0000,
        17051.5000, 17054.0000, 17054.0000, 17056.5000, 17061.5000, 17071.5000,
        17076.5000,  1730.5000,  1731.0000,  1736.0000,  1738.5000,  1738.5000,
         1745.5000,  1750.5000,  1750.5000, 17749.0000, 17752.0000, 17754.0000,
        17754.5000, 17756.5000, 17764.0000, 17766.5000, 17769.0000, 17769.5000,
        17771.5000, 17774.0000, 17774.5000, 17776.5000, 17779.0000, 17782.0000,
        17794.0000, 18429.5000, 18429.5000, 18437.0000, 18439.5000, 18442.0000,
        18444.0000, 18444.5000, 18444.5000, 18447.0000, 18449.0000, 18452.0000,
        18454.0000, 18454.0000, 18454.0000, 18456.5000, 18462.0000, 19087.0000,
        19147.0000, 19164.5000, 19174.5000, 19182.0000, 19182.0000, 19182.0000,
        19187.0000, 19189.5000, 19192.0000, 19194.5000, 19197.0000, 19199.5000,
        19202.0000, 19209.2000, 19789.5000, 19834.5000, 19842.0000, 19879.5000,
        19889.5000, 19892.0000, 19897.0000, 19897.0000, 19902.5000, 19910.0000,
        19917.5000, 19934.5000, 19962.5000, 20057.5000,  2393.0000,  2398.5000,
         2408.0000,  2411.0000,  2415.7000,  2420.5000,  2421.0000,  2425.5000,
         2428.5000,  2438.0000,  2443.0000,  3148.0000,  3148.5000,  3150.5000,
         3155.5000,  3156.0000,  3158.0000,  3163.0000,  3163.0000,  3163.1000,
         3163.5000,  3165.5000,  3173.0000,  3178.0000,  3850.5000,  3853.0000,
         3855.5000,  3858.0000,  3858.5000,  3860.5000,  3861.0000,  3863.0000,
         3868.0000,  3868.0000,  3868.0000,  3873.0000,  4523.0000,  4543.5000,
         4545.5000,  4548.5000,  4553.0000,  4553.0000,  4555.5000,  4555.5000,
         4555.5000,  4558.0000,  4558.0000,  4558.5000,  4560.5000,  4563.0000,
         4590.5000,  5355.5000,  5365.5000,  5368.0000,  5368.5000,  5370.5000,
         5370.5000,  5373.0000,  5378.0000,  5378.0000,  5383.0000,  5385.5000,
         5388.0000,  5390.5000,  5392.5000,  5397.5000,  5398.0000,  5408.5000,
         5422.5000,  6113.3000,  6115.5000,  6115.5000,  6115.5000,  6115.5000,
         6120.5000,  6123.0000,  6123.5000,  6123.5000,  6127.5000,  6127.8000,
         6130.5000,  6130.5000,  6133.0000,  6135.5000,  6137.5000,  6138.0000,
         6140.5000,  6761.0000,  6778.0000,  6780.5000,  6780.5000,  6780.5000,
         6783.0000,  6783.0000,  6785.5000,  6785.5000,  6788.0000,  6788.0000,
         6790.5000,  6790.5000,  6795.5000,  6797.5000,  6800.0000,  6800.5000,
         6825.5000,  7525.5000,  7528.0000,  7555.5000,  7558.0000,  7558.0000,
         7558.5000,  7560.5000,  7561.0000,  7568.0000,  7568.0000,  7570.0000,
         7570.5000,  7572.5000,  7573.0000,  7575.5000,  7575.5000,  7582.5000,
         7583.0000,  8283.0000,  8283.5000,  8285.5000,  8286.0000,  8287.5000,
         8288.0000,  8290.5000,  8290.5000,  8293.0000,  8295.5000,  8297.5000,
         8298.0000,  8302.5000,  8305.0000,  8305.5000,  8305.5000,  8308.0000,
         8315.0000,  8315.0000,  8325.5000,  8977.5000,  9025.0000,  9027.5000,
         9028.0000,  9028.5000,  9032.9000,  9033.0000,  9033.0000,  9033.5000,
         9035.5000,  9035.5000,  9038.5000,  9040.5000,  9040.5000,  9040.5000,
         9043.0000,  9043.0000,  9043.5000,  9048.0000,  9057.5000,  9060.5000,
         9736.0000,  9738.6000,  9748.5000,  9748.5000,  9750.5000,  9750.5000,
         9750.5000,  9753.0000,  9755.5000,  9760.5000,  9763.0000,  9763.5000,
         9765.5000,  9765.5000,  9765.5000,  9767.5000,  9770.5000,  9772.5000,
         9775.5000,  9775.5000,  9787.5000], dtype=torch.float64), tensor([ 7065.5000,  5607.5000,  2705.0000,  1242.5000,  6352.5000,  7805.5000,
         9260.5000,  4163.5000, 15780.5000, 13630.5000,  4897.5000, 10003.0000,
         8563.0000,  1985.0000,  3435.0000, 12185.5000, 10735.5000, 12903.0000,
        14283.0000, 11455.5000,  4910.5000,  4235.0000,  9980.5000,  1972.5000,
         3407.5000,  4872.5000,  4175.0000,  1225.0000, 11445.5000, 12158.0000,
         6330.5000,  8538.0000,  7805.5000,  9253.0000,  2697.5000,  5597.5000,
        15770.5000,  7072.5000, 10728.0000, 15033.0000, 14335.5000, 12890.5000,
        13613.0000,  8525.5000,  6332.5000, 10725.5000,  1960.0000,  2702.5000,
         3417.5000, 15045.5000,  7795.6000,  1220.0000,  4875.0000,  5612.5000,
         9268.0000, 10005.5000, 11433.0000,  4160.0000,  7087.5000, 12898.0000,
        14335.5000, 12165.5000, 13623.0000, 13698.0000,  2692.5000,  4135.0000,
         6302.5000, 11418.0000, 14313.0000,  1932.5000,  3417.5000,  9975.5000,
        10698.0000, 15025.5000,  9248.0000,  7785.5000, 12153.0000,  7062.5000,
         8510.5000, 12870.5000,  5585.0000,  4860.0000,  4905.0000,  7072.5000,
         5632.5000, 11468.0000, 10740.5000,  3450.0000,  7810.5000,  8553.0000,
        15061.0000,  4165.0000,  1982.5000,  6332.5000,  9273.0000, 13630.5000,
         2720.0000, 12170.5000, 14355.5000,  9995.5000,  7027.5000, 11465.5000,
         1262.5000,  3457.5000,  4167.5000,  8565.5000, 10745.5000,  2717.5000,
         4917.5000,  5650.0000,  7822.5000, 14365.5000, 15078.5000,  6360.0000,
         1992.5000, 12925.5000, 10013.0000, 12195.5000,  9290.5000, 13648.0000,
         7822.5000,  8535.5000,  6350.0000,  3440.0000, 10718.0000,  1984.5000,
        12918.0000,  9288.0000,  5627.5000,  9998.0000, 11463.0000,  7092.5000,
         4170.0000, 12180.5000, 13633.5000, 14343.5000,  4905.0000,  2702.5000,
         9993.0000, 13635.5000,  2717.5000,  5625.0000,  1982.0000,  4890.0000,
        11453.0000,  9280.5000,  8547.8000, 14353.5000,  7817.5000,  7095.0000,
         3430.0000,  4167.5000,  6357.5000, 12915.5000, 12183.0000, 10765.5000,
         9988.0000,  7820.0000, 12893.0000,  1974.5000,  7095.0000,  4167.5000,
         9270.5000,  2709.5000,  3442.5000,  4885.0000,  5615.0000,  6357.5000,
         8540.5000, 11463.0000, 10735.5000, 14351.0000, 12193.0000,  2782.0000,
         4170.0000,  3427.0000,  9253.0000, 10730.5000,  6342.5000,  8535.5000,
         7812.5000,  4880.0000,  5615.0000,  7075.0000, 11458.0000, 12195.5000,
         9990.5000,  9285.5000,  4945.5000,  5635.5000,  6373.0000,  7110.5000,
        10025.5000,  7843.0000,  8563.0000,  7825.0000,  4912.5000, 10013.0000,
         3449.5000,  6350.0000,  8537.5000,  9270.5000, 11465.5000,  5632.5000,
        13638.5000, 10740.5000,  4169.5000, 12193.0000,  7067.5000,  2704.5000,
        14353.5000,  5620.0000,  7057.5000,  2684.5000,  8520.5000,  7805.0000,
        13626.0000,  4157.0000,  9275.5000,  4892.5000,  9978.0000,  3432.0000,
        11448.0000, 12901.0000, 10715.5000, 12170.5000,  6342.5000,  3424.5000,
         4142.0000,  7060.0000,  6340.0000, 12165.5000,  5582.5000,  9260.5000,
         7802.5000,  8522.5000,  4872.5000,  9973.0000, 11438.0000, 10715.5000,
        12893.5000, 13618.6000, 10703.0000,  9968.0000,  7860.0000,  9250.5000,
         8512.5000, 12163.0000, 11448.0000, 12891.0000,  4894.5000,  5592.5000,
         6332.5000, 13543.5000,  4112.0000,  7007.5000, 11495.5000,  6360.5000,
         7845.5000,  4180.5000,  5640.4000, 10733.0000,  4913.0000,  8573.0000,
         7095.5000,  9985.5000,  9278.0000,  7808.0000,  4173.0000,  4883.0000,
         7078.0000,  3440.5000,  8560.5000, 12178.0000,  9995.5000,  6360.6000,
         5625.5000, 10730.5000, 11460.5000,  9273.0000,  8543.0000,  9275.5000,
         5620.5000, 12183.0000,  4170.5000,  7065.5000,  3440.5000, 10013.0000,
        11458.0000, 10733.0000,  6360.5000,  7803.0000,  6420.5000,  2695.5000,
        12893.0000,  3425.5000,  7060.5000,  8520.5000, 10723.0000,  7795.5000,
         9253.0000,  5605.5000,  9988.0000,  4155.5000, 12180.5000,  4870.5000,
        11453.0000,  4950.5000,  8553.0000,  7838.0000,  1270.5000,  2710.5000,
         3463.0000,  7093.0000, 10030.5000,  9275.5000,  4183.0000,  5618.0000,
        10730.5000,  6368.0000, 12208.0000, 13643.0000, 11460.5000,  1998.0000,
        12925.5000,  3473.1000,  2728.0000,  4918.0000,  7110.5000,  9270.5000,
         7848.0000,  5638.0000,  1268.0000,  2013.0000, 13658.0000, 11475.4000,
        10018.0000, 12220.5000,  6385.5000, 10765.5000, 12935.5000,  4200.5000,
         8575.5000,  1270.5000, 10020.5000, 10755.5000, 11473.0000,  1985.5000,
         4905.5000,  8575.5000,  2728.0000,  7820.5000,  7095.5000,  9278.0000,
         3443.0000,  5645.5000,  4205.5000, 12923.0000, 13628.0000, 12200.5000,
         6330.5000,  2710.5000,  7113.0000,  4193.0000,  6360.5000,  3450.5000,
         1995.5000,  7830.5000,  1258.0000, 10750.5000,  4923.0000, 13655.5000,
         9285.5000, 12208.0000,  8563.0000, 10030.5000, 11470.5000, 12925.5000,
         5603.0000,  8550.5000,  1990.5000,  2715.5000,  1255.5000, 12190.5000,
         4168.0000,  3450.5000,  9288.0000,  5643.0000, 10728.0000, 12918.0000,
         4910.5000, 15785.5000, 14353.0000, 11478.0000,  7110.5000, 10020.5000,
        13640.5000, 15073.0000,  6355.5000, 15125.5000, 14330.5000, 13638.0000,
        10723.0000,  1970.0000,  5627.8000,  6360.5000,  9253.0000,  1245.0000,
        12915.5000,  4165.5000,  3450.5000, 12178.0000,  4902.5000,  7823.0000,
         8545.5000,  9988.0000,  2717.5000,  7095.5000, 15788.0000, 11468.0000,
         1992.5000,  4190.1000,  1265.0000,  3480.0000,  5635.5000,  7095.5000,
         7840.5000, 11478.0000,  8575.5000,  4907.5000, 12923.0000,  2735.0000,
        10008.0000, 12205.5000,  9278.0000, 14355.5000,  6370.5000, 15798.0000,
        10750.5000, 13635.5000, 15093.0000], dtype=torch.float64)]]

3.2 Get coordinates of sub-spots from within spots

We use reshape_latent_image() function to convert the super-resolved gene expression with 16X relolution reconstructed_matrix_reshaped (torch.Size([21296, 596])) into tensor format reconstructed_matrix_reshaped_tensor (torch.Size([1331, 16, 596])).

[50]:

reconstructed_matrix_reshaped_tensor, _ = fst.reshape_latent_image(reconstructed_matrix_reshaped,
                                                                   dataset_class='Visium16')
print("The size of reconstructed tensor data:", reconstructed_matrix_reshaped_tensor.shape)

The size of reconstructed tensor data: torch.Size([441, 16, 647])

[53]:

# scale_factors = fst.json_load(f"{path}/FineST/FineST_local/Dataset/NPC/patient1/")
# print(scale_factors['spot_diameter_fullres'])

[51]:

(first_spot_first_variable, C,
 _, _, _) = fst.subspot_coord_expr_adata(reconstructed_matrix_reshaped_tensor, adata, gene_hv,
                                         p=0, q=0, patch_size=56, dataset_class='Visium16')
print("first_spot_first_variable shape:", first_spot_first_variable.shape)

pixel_step (half of patch_size): 7.0
first_spot_first_variable shape: (16,)

[52]:

fstplt.subspot_expr(C, first_spot_first_variable, patch_size=56, dataset_class='Visium16',
                     marker='s', s=1050, rotation=30, save_path=None)
_images/Demo_Train_Impute_count_81_0.png 
[53]:

## Get the sub-spot level gene expression of all genes in all within spots
(_, _, all_spot_all_variable,
 C2, adata_infer) = fst.subspot_coord_expr_adata(reconstructed_matrix_reshaped_tensor,
                                                adata, gene_hv, patch_size=56, dataset_class='Visium16')
print("all_spot_all_variable shape:", all_spot_all_variable.shape)
print(adata_infer)

pixel_step (half of patch_size): 7.0
all_spot_all_variable shape: (7056, 647)
AnnData object with n_obs × n_vars = 7056 × 647
    obs: 'x', 'y'
    obsm: 'spatial'

[54]:

print(adata_infer.obs[['x', 'y']])
print(adata_infer.obsm['spatial'])
print(adata_infer.obsm['spatial'][:,0].max(), adata_infer.obsm['spatial'][:,1].max())

          x      y
0     10394   7044
1     10408   7044
2     10422   7044
3     10436   7044
4     10394   7058
...     ...    ...
7051   9808  15100
7052   9766  15114
7053   9780  15114
7054   9794  15114
7055   9808  15114

[7056 rows x 2 columns]
[[10394  7044]
 [10408  7044]
 [10422  7044]
 ...
 [ 9780 15114]
 [ 9794 15114]
 [ 9808 15114]]
20078 15819

Here, adata_infer is the inferred super-resolved gene expression data with 16X solution.

[55]:

_, adata_infer_reshape = fst.reshape_latent_image(torch.tensor(adata_infer.X), dataset_class='Visium16')
print("adata_infer shape:", adata_infer.shape)
print("adata_infer_reshape shape:", adata_infer_reshape.shape)

adata_infer shape: (7056, 647)
adata_infer_reshape shape: torch.Size([441, 647])

[56]:

fstplt.gene_expr_compare(adata, "CD27", adata_infer_reshape, gene_hv, s=250, save_path=None)
_images/Demo_Train_Impute_count_86_0.png 
[57]:

fstplt.gene_expr_compare(adata, "ERBB2", adata_infer_reshape, gene_hv, s=250, save_path=None)
_images/Demo_Train_Impute_count_87_0.png 
[62]:

fstplt.gene_expr(adata, adata_infer_reshape, gene_selet='ERBB2', marker='o',
                 s=80, figsize=(5, 4), cnt_color='turbo',
                 # save_path=None)
                 save_path='ERBB2_expr_infer.pdf')

fstplt.gene_expr(adata, adata_infer_reshape, gene_selet='CD27', marker='o',
                 s=80, figsize=(5, 4), cnt_color='turbo', save_path='CD27_expr_infer.pdf')
_images/Demo_Train_Impute_count_88_0.png
_images/Demo_Train_Impute_count_88_1.png 
[60]:

fstplt.mean_cor_box(adata, adata_infer_reshape, save_path='Boxplot_infer_cor_count.pdf')
fstplt.mean_cor_box(adata_norma, adata_infer_reshape, save_path='Boxplot_infer_cor_norml.pdf')

0.8438119241131888
0.518594758845764
_images/Demo_Train_Impute_count_89_1.png 
0.7469443256238438
0.37708428803191435
_images/Demo_Train_Impute_count_89_3.png 
[56]:

adata_infer_spot = fst.reshape2adata(adata, adata_infer_reshape, gene_hv)
adata_infer_spot.to_df()

[56]:
TGFB1 TGFBR1 TGFBR2 TGFB2 TGFB3 ACVR1B ACVR1C ACVR1 BMP2 BMPR1A ... SEMA6D KDR TREM2 SEMA6A SEMA6B SEMA7A PLXNC1 SIGLEC1 THY1 VCAM1
15x19 0.081259 0.016418 0.010612 0.007467 0.016015 0.007447 0.008951 0.022685 0.012328 0.008821 ... 0.008332 0.007789 0.016270 0.009386 0.008195 0.013419 0.007877 0.007938 0.029722 0.008144
15x21 0.013452 0.026958 0.007591 0.005829 0.007995 0.007785 0.008139 0.012783 0.008177 0.008930 ... 0.007441 0.006457 0.007581 0.008397 0.028308 0.006756 0.007311 0.008409 0.019908 0.006688
15x25 0.032207 0.159735 0.038286 0.009353 0.007072 0.006671 0.007533 0.033302 0.014748 0.030480 ... 0.007550 0.007322 0.027524 0.008183 0.007379 0.006517 0.010280 0.008172 0.216861 0.007843
15x27 0.049247 0.024144 0.008317 0.006575 0.011221 0.008214 0.008096 0.027744 0.009017 0.018914 ... 0.007882 0.007843 0.013813 0.007703 0.014229 0.018180 0.007720 0.013550 0.162597 0.007971
15x20 0.119307 0.048920 0.011436 0.007691 0.011840 0.014636 0.007692 0.008293 0.008435 0.010631 ... 0.006989 0.043717 0.011705 0.007852 0.012633 0.013552 0.015486 0.009103 0.386875 0.012834
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
16x20 0.051323 0.041818 0.027178 0.007682 0.012097 0.010532 0.013264 0.016291 0.012213 0.024528 ... 0.008085 0.007328 0.008270 0.008015 0.008215 0.017346 0.018824 0.008693 0.176034 0.014691
16x7 0.052623 0.008779 0.008794 0.006862 0.008059 0.007028 0.008790 0.008925 0.008475 0.008356 ... 0.007908 0.008506 0.008478 0.008684 0.009374 0.007993 0.006814 0.009024 0.065491 0.008382
16x14 0.019976 0.008926 0.008053 0.006833 0.007763 0.007272 0.014728 0.009765 0.069536 0.008028 ... 0.008074 0.008174 0.009039 0.007924 0.008372 0.007692 0.006845 0.008218 0.008627 0.008122
16x10 0.017288 0.026855 0.037522 0.005255 0.007230 0.007821 0.013074 0.178009 0.006572 0.007531 ... 0.007216 0.005949 0.005366 0.052425 0.029245 0.007041 0.005913 0.007046 0.016135 0.006503
16x8 0.041181 0.008730 0.008442 0.006769 0.007949 0.006879 0.011146 0.008804 0.008513 0.008530 ... 0.007691 0.008616 0.008805 0.010071 0.021671 0.007657 0.006037 0.008802 0.056768 0.008645

441 rows × 647 columns

[63]:

fstplt.cor_hist(adata, adata_infer_spot.to_df(), max_step=0.2, min_step=0.05,
                 fig_size=(5, 4), trans=False, format='svg',
                 save_path='Hist_infer_cor_count_allLRgs.svg')

nan
_images/Demo_Train_Impute_count_91_1.png 
[64]:

print(adata_infer)
print(adata_infer_spot)

AnnData object with n_obs × n_vars = 7056 × 647
    obs: 'x', 'y'
    obsm: 'spatial'
AnnData object with n_obs × n_vars = 441 × 647
    obs: 'in_tissue', 'array_row', 'array_col'
    var: 'mt', 'n_cells'
    obsm: 'spatial'

[65]:

## save adata 85184 × 596
patientxy = 'Demo_FineST'
adata_infer.write_h5ad(str(path)+'FineST/FineST_local/Dataset/ImputData/'+str(patientxy)+'/'+str(patientxy)+'_adata_infer.h5ad')

[66]:

## save adata 1331 × 596
patientxy = 'Demo_FineST'
adata_infer_spot.write_h5ad(str(path)+'FineST/FineST_local/Dataset/ImputData/'+str(patientxy)+'/'+str(patientxy)+'_adata_infer_spot.h5ad')

[63]:

pwd

[63]:

'/mnt/lingyu/nfs_share2/Python/FineST/FineST_local/Dataset/ImputData/Demo_FineST/Figure'

[60]:

gene='ERBB2'
import scipy.stats as st
print(st.pearsonr(adata_infer_spot.to_df()[str(gene)].values, adata_infer_spot.to_df()[str(gene)].values))

PearsonRResult(statistic=0.9999999999999998, pvalue=0.0)