Darvesh Gorhe

MS Data Science @ Columbia University (2022)

I'm a recent graduate student from Columbia University's the Master's of Science in Data Science (MSDS) program. I'm currently looking for a full-time job so if you are looking for a bioinformatics scientist, computational biologist, or data scientist, feel free to contact me through any of the links on this page. I did a computational biology internship at Roche Diagnostics and I'm currently working as a Research Assistant at the Jovanovic Lab at Columbia University. I try to share all my code/learnings on GitHub. However, I'm not able to share code related to ongoing papers/projects in the Jovanovic Lab or code related to projects I worked on at Roche Diagnostics.

SECAT Performance Improvements

Improved Size-exclusion chromatography toolkit (SECAT) by ~40% through optimizations and multiprocessing support for non-parametric Loess normalization.

Kitchen Pantry

(In progress) application to help users keep track of food and recipes.

Personal Blog

A space where I apply my data science frameworks and skills to contemporary questions.

Research Assistant | Jan 2022 — Current

Jovanovic Lab @ Columbia University

• Working with protein turnover data to identify differentially expressed proteins and differential complex turnover in primary versus metastatic pancreatic cancer cells using Python with the following packages: bokeh, networkx, scipy, pandas, matplotlib, numpy.
• Optimized Size-Exclusion Chromatography Toolkit (SECAT) by reducing memory intense data processing and adding asynchronous conconcurrency with Python multiprocessing package, and improved the user experience and developer experience via conda environments.
• Performing outlier analysis for differentially phosphorylated proteins in primary pancreatic cancer cells versus metastatic liver cancer cells using R and Python with the following packages: blacksheep, gseapy, scipy, pandas, numpy, and ssGSEA2.0.

Computational Biology Intern | May 2022 — Aug 2022

Roche Molecular Diagnostics, Research and Early Development

• Implemented convex optimization based gene prioritization with PyTorch neural network module to explore novel polygenic risk scoring for Alzheimer’s Disease using 3 data modalities. Worked on a multi-discplinary approach using molecular biology, computer science, and data science.
• Independently developed graph neural network classifier to determine whether an individual has Alzheimer’s Disease based on heterogenous graph with 4 node modalities and 3 edge modalities.
• Created machine learning operations (ML Ops) infrastructure to automate graph neural network training, validation, testing, and inference via PyTorch Lightning.

Capstone Project Member | Sep 2022 — Dec 2022

Columbia University (Data Science Institute)

• Transformed Jupyter notebooks for scRNA preprocessing and analysis into Python scripts to improve the repro- ducibility and scalability of NGS analysis.
• Created Docker containers for CellPhoneDB, Tangram, and SCRAN/Liger analyses to improve the portability of the NGS analyses across Windows, Linux Servers, and MacOS.
• Performed CPU and memory profiling via scalene and used insights to make several memory and file I/O improvements including: vectorizing counts normalization code to improve speed by 98.1%, switching from csv to parquet file format re- ducing file size by up to 97%, and dynamically switching between regular and compressed sparse row (CSR) representations based on sparsity of counts matrices.