Gene-compression-and-Cancer-type-classification

Fall 2021 Final Project for CM226 - Machine Learning in Bioinformatics

Dataset

• TCGA - The Cancer Genome Atlas PanCanAtlas RNAseq data from the National Cancer Institute Genomic Data Commons
• These data consisted of 11,069 samples with 20,531 measured genes. Preprocessing-
• Tumors that were measured from multiple sites were removed.
• Data was normalised
• This resulted in a final TCGA PanCanAtlas gene expression matrix with 11,060 samples, which included 33 different cancer types, and 16,148 genes.
• The data is split into 90% training and 10% testing partitions. The data is partitioned such that each split contained relatively equal representation of each cancer type.

Proposed Model

Implementation of PCA, NMF, ICA

• These models have been built using sklearn library

Implementation of VAE Model

• This VAE model is inspired from Tybalt's implementation

Authors

• Rushi Bhatt
• Ronak Kaoshik
• Shruti Mohanty

See also the list of contributors who participated in this project.

License

This project is licensed under the MIT License - see the LICENSE.md file for details