CS6024-Algorithmic-Approaches-to-Computational-Biology
This repository contains the code and assignment solutions to the CS6024- algorithmic approaches to computational biology course.
It consists of the 5 assignments given in the course. As well as the semester long project documentation. The project was on extracting a biologically relevant latent space from cancer gene expression data using deep variational autoencoders. The learned embeddings were further studied through t-SNE and hierarchal clustering dendograms visualizations. The Encoder of the VAE was used to perform downstream pan-cancer (33 subtypes) and organ classification. We also used model interpretability tools such as SHAP to map the relevant genes to each cancer subtype and organ. We were able to show that the embeddings learned by the model gave the same gene-target mapping as seen in literature.