GSOC 2025 Project: Integrating Single Cell RNA-seq modeling for Cell Type Identification into Deepchem

Community
1

Hi everyone!

I’m Harindhar, a final-year Dual Degree student in Biological Engineering at IIT Madras. I’m using this thread to document my GSoC 2025 project with DeepChem. As part of the project, I’ll be integrating ACTINN, a neural network model for automated cell type identification in single-cell RNA-seq data, into the DeepChem ecosystem.

I’ll be sharing weekly updates here throughout the summer as the project progresses. Stay tuned!

2 Likes
GSoC ‘25 | Integrating Single Cell RNA-seq modeling into Deepchem | Final Report
2

Progress June 2 - 8

Ran training and evaluation loops for the existing TF and PyTorch implementations.

Roadblocks: Script kept crashing during data manipulation (normalization, concatenation, etc.) due to dataset size (20k genes × 50k cells), though it loaded fine in pandas.

Fixes: Sampled 1k cells to run the code. Sparse genes led to zero total expression, causing NaNs when computing CV (std/mean). Filtered out genes with zero mean.

To do:

  • Create a DeepChem iterator using single cell rna-seq data
  • Add the classification layer
3

Progress: June 9–16

  • Converted the .h5 file to a .csv format and sampled 1,000 rows due to memory limitations. Created a DiskDataset object using CSVLoader.
  • Currently working on a featurizer to normalize and filter genes, targeting the bottom and top 1%.
  • Exploring alternative normalization and gene filtering strategies, since the full dataset cannot be loaded into memory at once.
4

Week 3 (June 16 - 22)

  • Draft PR to merge the featurizer for ACTINN (PR#4463)
  • Works with DeepChem’s CSVLoader class

To Do:

  • Figure out how to process the data (the implementation provided with the paper processes train and test data together)
5

Week 4 (June 23 - 29)

  • Made changes to the draft PR - changed the feature selection from using both train and test to using only the train set
  • Wrote a unit test for the featurizer
  • Working on dataloader, layer, and model classes
6

Week 5 (June 30 - July 6)

  • Implemented code to accumulate total expression across shards and filter genes afterward. Integration with DeepChem is still pending.

  • Completed implementation of the custom layer class

  • Currently developing the model class

7

Week 6 (July 7 - July 13)

8

Week 7 (July 14 - July 20)

  • Implemented the prototype for the model in deepchem style, trained it, replicated an experiment from the paper and got a similar accuracy. https://github.com/deepchem/deepchem/pull/4487
  • Working on the tutorial for the model
  • Implemented the data transformations as a deepchem Transformer subclass instead of a Featurizer subclass
9

Week 8

  • The tutorial for training the model is ready (PR #4496).
  • Made fixes to the tutorial based on mentor feedback
  • I’ve also been spending time understanding DeepChem and exploring how scRNA-seq models could be integrated.
10

Week 9 (July 26 - Aug 1)

  • Made updates to the ACTINN tutorial (PR #4496) based on mentor feedback, including more background on scRNA-seq and preprocessing.
  • Currently working on extending the Scanpy tutorial present in Deepchem
11

Week 10 (Aug 2 - Aug 8)

  • Updated the ACTINN tutorial to include more explanation about single-cell RNA sequencing, the tools used, and common analysis methods.(PR#4496)
  • Improved the Scanpy tutorial by showing how to use its results to train and test a model with DeepChem.(PR#4506)
12

Week 11 (Aug 9- Aug 15)

  • Improved the Scanpy tutorial by showing how to use its results to train and test a model with Deepchem.(PR#4506).
  • Worked on deepchemizing the preprocessing pipeline.
  • Spent time understanding Deepchem conventions to adapt preprocessing functions accordingly.
13

Week 12 (Aug 16- Aug 22)

  • Merged PR extending Scanpy tutorial by adding deepchem support (PR#4506, PR#4520)
  • Worked on deepchemizing the preprocessing pipeline
  • Spent time understanding preprocessing in scRNA seq data analysis