OmixHub Documentation

Overview

OmixHub is a platform that interfaces with GDC using Python to help users apply ML-based analysis on different sequencing data. Currently, we support only RNA-Seq based datasets from the genomic data commons (GDC).

• Cohort Creation of Bulk RNA Seq Tumor and Normal Samples from GDC.
• Bioinformatics analysis: Application of PyDESeq2 and GSEA in a single pipeline.
• Classical ML analysis: Applying clustering, supervised ML, and outlier sum statistics.
• Custom API Connections:
  • Search and retrieval of Cancer Data cohorts from GDC using complex JSON filters.
  • Interacting with MongoDB in a Pythonic manner (DOCS coming soon).
  • Interacting with Google Cloud BigQuery in a Pythonic manner (DOCS coming soon).

Getting Started

1. Clone the repository:

   git clone https://github.com/adhal007/OmixHub.git

2. Create the correct conda environment for OmixHub:

   conda env create -f environment.yaml

API Reference

For detailed API reference, please visit the full documentation:

Examples

Useful query filters for GDC API endpoints

For examples of using various filters from the GDCQueryFilters class, please refer to the examples section in the full documentation.

Data Processing and Analysis Examples

• Cohort Creation of Bulk RNA Seq Experiments from GDC
• Migrating GDC RNA-Seq Expression Data to BigQuery
• Differential Gene Expression and GSEA Analysis
• Supervised ML on a dataset from BigQuery

Applications

Running Gradio App

Currently restricted to users. To contribute or try it, contact adhalbiophysics@gmail.com.

Running the app:

python3 app_gradio.py

For app navigation documentation, refer to: cd ../tutorial_notebooks/docs/UI Prototype/gradio_use.md

References

1. Characterizing tumor toxicity in Gene therapy targets from Bulk RNA-Sequencing
2. Bayesian Framework for identifying gene expression outliers in individual sample of RNA-Seq data

Additional Documentation

• Application Examples
  • Outlier sum statistics
  • Supervised ML using Ensemble models on kidney cancer data
  • Clustering using bayesian optimized parameters of kidney cancer sub-types from TCGA
  • Analysis of differentially expressed genes functionally
• Roadmap for future developments
• Methods in src.Connectors for GDC API search and retrieval using custom queries