Tutoring on usage of HuggingFace transform model for genomics
Budget: $30 – $250 USD
Title: Tutoring on HuggingFace transform model for genomics
I am looking for an expert to help me use a publicly available transformer model, Geneformer, available here: https://huggingface.co/ctheodoris/Geneformer.
Model details described here: https://www.nature.com/articles/s41586-023-06139-9.
Example code for the tasks of interest available here: https://huggingface.co/ctheodoris/Geneformer/tree/main/examples
At the end of this project I would like to be able to perform the following tasks:
- Preprocess new data as required for use with the model
- Fine-tune model with new data for applications of interest (described below)
- Utilize model for the following tasks:
o Obtain cell type specific hierarchical gene regulatory networks,
o Perform in silico deletion or activation of genes for disease modeling and treatment effect predictions,
o Cell type classification of new scRNA-seq samples,
o Bonus: batch correction of scRNA-seq data
Deliverables for this project are:
- python (or R) notebooks that can achieve the tasks above starting from 10x genomics count data (raw or filtered barcodes.tsv.gz, features.tsv.gz, matrix.mtx.gz - other possible starting files would be raw or filtered_feature_bc_matrix.h5 or cloupe.cloupe).
- Bonus: code to also preprocess bulk RNAseq data for use with the model (from count matrix of genes by samples).
- Bonus: Dockerfile with all necessary dependencies to run the code, extra bonus if this works on AWS SageMaker Studio.
I am looking for an expert to help me use a publicly available transformer model, Geneformer, available here: https://huggingface.co/ctheodoris/Geneformer.
Model details described here: https://www.nature.com/articles/s41586-023-06139-9.
Example code for the tasks of interest available here: https://huggingface.co/ctheodoris/Geneformer/tree/main/examples
At the end of this project I would like to be able to perform the following tasks:
- Preprocess new data as required for use with the model
- Fine-tune model with new data for applications of interest (described below)
- Utilize model for the following tasks:
o Obtain cell type specific hierarchical gene regulatory networks,
o Perform in silico deletion or activation of genes for disease modeling and treatment effect predictions,
o Cell type classification of new scRNA-seq samples,
o Bonus: batch correction of scRNA-seq data
Deliverables for this project are:
- python (or R) notebooks that can achieve the tasks above starting from 10x genomics count data (raw or filtered barcodes.tsv.gz, features.tsv.gz, matrix.mtx.gz - other possible starting files would be raw or filtered_feature_bc_matrix.h5 or cloupe.cloupe).
- Bonus: code to also preprocess bulk RNAseq data for use with the model (from count matrix of genes by samples).
- Bonus: Dockerfile with all necessary dependencies to run the code, extra bonus if this works on AWS SageMaker Studio.