4i_analysis

A set of notebooks to align 4i data coming from a Nikon widefield.

Analysis flow:

General idea:

• in each round an image that is easy for segmentation is selected
• these images are segmented using Cellpose
• masks are aligned using StackReg and transformations are saved
• transformations are applied to all channels
• aligned images are stored as single channels tiffs

00_preprocess_data.ipynb

• creates data frames with the info for a given well (ex. df)
• saves images prepared for the segmentation (ex. im_to_segment)
• output directories will be created automatically (if their parent directories exist)

01_cellpose_segmentation_local.ipynb (01_cellpose_segmentation_colab.ipynb)

• uses Cellpose segmentation
• segments and saves masks (ex. im_to_segment)
• this segmentation is performed on the local workstation
• output directory will be created automatically (if its parent directory exist)

02_find_transforms_on_segmentation_df.ipynb

• calculate transformations to align segmented images
• allows calculation of transformations on downscaled images
• output directory will be created automatically (if its parent directory exist)

03_find_transformation_manually.ipynb

• uses Affinder to find a transformation a the problematic well/round
• replaces transformation in the pkl file

04_align_from_transform_single.ipynb (04_align_from_transform_list.ipynb)

• applies transformations to all original images
• works for a single well
• allows visualization of alignment
• output directory will be created automatically (if its parent directory exist)

05_create_mask.ipynb

• enables creation of punchmask (a mask of regions to be excluded from the analysis because of the artefacts)
• ex. output subdirectory - masks
• output directory will be created automatically (if its parent directory exist)

06_calculate_cell_properties.ipynb

• calculates properties of individual segmented nuclei through all the intensity channels
• ex. output subdirectory - cell_data
• output directory will be created automatically (if its parent directory exist)

Alternatively:

01_cellpose_segmentation_colab.ipynb

• segments and saves masks (ex. im_to_segment)
• this notebook is adjusted to be run on Google Colab (while data are temporarily stored on Google Drive)
• the advantage is access to GPU which will usually speed up segmentation 5-10 times
• output directory will be created automatically (if its parent directory exist)

04_align_from_transform_list.ipynb

• enables processing of a list of wells
• doesn't support visualization of the alignment
• output directory will be created automatically (if its parent directory exist)

Starting data structure:

• a set of correctly stitched nd2 files
• these stitched images may be of slightly different sizes (as stitching in Elements software corrects for the imperfections of a microscope stage position). However, if the sizes are totally different, please check your dataset.
• images are divided into subdirectories according to rounds (names of subdirectories contain names of the rounds in a form: 'Round_ROUNDNAME_WHATEVERADDITIONALINFO' )
• please make sure that the names of the rounds are unique (ex. no subdirectories as 'Round_00_bad' and 'Round_00_good')
• keep names of the rounds numeric
• assuming standard names of stitched nd2 files (starting with 'WellA1_Channel...', where 'A1' is the well name)
• you can keep individual files of tiles in the same subdirectories as they will be ignored (tiled images usually have names starting with 'WellA1_Point...')

Info file:

• a csv file that translates names of optical configurations (OC) used in specific rounds into desired names for the channels (ex. names of the antibodies)
• for example:

myRound	Hoechst	AF488	AF555	AF647
00	DNA	pRb		RB
01	DNA	p27	p53

• where 'Hoechst', 'AF488' and 'AF647' are the names of the optical configurations (OC)
• it is ok to leave the spots empty if a given OC was not used in some rounds

Example input data structure:

• data
  • Round_00_DNA_pBB_RB
    • WellA1_Channel___.nd2
    • WellA2_Channel___.nd2
  • Round_01_DNA_p27_p53
    • WellA1_Channel___.nd2
    • WellA2_Channel___.nd2
• info_exp.csv

Example output data structure:

• data
• df
  • df_A1.pkl
  • df_A2.pkl
  • tmat_A1.pkl
  • tmat_A2.pkl
• im_to_segment
  • A1
    • im2segment_A1_round_00.tif
    • im2segment_A1_round_01.tif
  • A2
    • im2segment_A2_round_00.tif
    • im2segment_A2_round_01.tif
• im_segmented
  • A1
    • labels_A1_round_00.tif
    • labels_A1_round_01.tif
  • A2
    • labels_A2_round_00.tif
    • labels_A2_round_01.tif
• aligned_tiffs
  • A1
    • Round_00_wellA1_DNA.tif
    • Round_00_wellA1_pRB.tif
    • Round_00_wellA1_RB.tif
    • Round_01_wellA1_DNA.tif
    • Round_01_wellA1_p27.tif
    • Round_01_wellA1_p53.tif
  • A2
    • Round_00_wellA2_DNA.tif
    • Round_00_wellA2_pRB.tif
    • Round_00_wellA2_RB.tif
    • Round_01_wellA2_DNA.tif
    • Round_01_wellA2_p27.tif
    • Round_01_wellA2_p53.tif
• masks
  • mask_A1.pkl
  • mask_A2.pkl
• cell_data
  • cell_data_A1.pkl
  • cell_data_A2.pkl
• info_exp.csv