Smriti

Smriti Vaidyanathan

I'm a Master's student in Computer Science at Columbia University, conducting research at the New York Genome Center in the Knowles Lab. My work focuses on developing deep learning models for integrative analysis of gene expression and alternative splicing in single-cell transcriptomic data.

Previously, I earned my Sc.B. in Computational Biology from Brown University, where I studied alternative splicing regulation by RNA-binding protein condensates using computational analysis of bioimages in the Larschan Lab.

Email GitHub LinkedIn

Publications

Robust Integration of Sparse Single-Cell Alternative Splicing and Gene Expression Data with SpliceVI

Pre-Print (2025)

Smriti Vaidyanathan*, Keren Isaev*, Aaron Zweig, David A Knowles

A splicing-aware multimodal VAE that jointly models AS and GE to produce more informative cell embeddings and uncover unique biological differences.

VAEs Single-Cell Representation Learning Multimodal Integration

Dual DNA/RNA-binding factor regulates dynamics of hnRNP splicing condensates

Pre-Print (2024)

Mukulika Ray, Julia Zaborowsky, Pranav Mahableshwarkar, Smriti Vaidyanathan, Jasmine Shum, Renjith Viswanathan, Annie Huang, wh Szu-Huan Wang, Victoria Johnson, Noah Wake, Ashley M. Conard, Alexander E. Conicella, Ryan Puterbaugh, Nicolas L. Fawzi, Erica Larschan

Research on the regulatory mechanisms of hnRNP splicing condensates and their dynamics, exploring how dual DNA/RNA-binding factors influence alternative splicing processes.

RNA Biology Splicing Phase Condensates Molecular Biology

Sex-specific splicing occurs genome-wide during early Drosophila embryogenesis

eLife (2023)

Mukulika Ray, Ashley Mae Conard, Jennifer Urban, Pranav Mahableshwarkar, Joseph Aguilera, Annie Huang, Smriti Vaidyanathan, Erica Larschan

Comprehensive analysis revealing widespread sex-specific alternative splicing patterns during early Drosophila development, providing insights into sex determination mechanisms.

Genomics Splicing Drosophila Developmental Biology

Projects

SpliceVI

Research Project within Knowles Lab: Multimodal VAE to integrate gene expression and splicing signals from scRNA-seq data (see preprint above), built using scvi-tools, inspired by and adapted from MultiVI. Also includes a single-modality splicing VAE built with a partial encoder and linear decoder.

Deep Learning Single-cell Python Splicing

iSTAR-C

Final project for Columbia's BMCS E4480: Extended iSTAR (Zhang et al., 2024) with biologically informed supervision for pixel-level cell-type prediction, using Starfysh-derived spot-level compositions and optional heat-diffusion pseudo-labels for out-of-spot supervision.

Deep Learning Spatial Transcriptomics Python Histology

Remembering to Learn: A Brain-Inspired Approach to Catastrophic Forgetting in AI

Final project for Columbia's COMS E6998 005: Extended FlyModel with BI-R, EWC, and cascade synapses to improve continual learning on CIFAR-100. Cascade synapses notably reduced memory loss without replay.

Continual Learning Neural Networks CIFAR-100 Python

Brain Tumor Segmentation and Survival Prediction (BRATS 2020)

Final project for Columbia's COMS W4995 005: Deep Learning for Computer Vision. Segmented tumor regions and predicted survival outcomes from BRATS 2020 MRI scans using U-Net and ResNet-based models.

Medical Imaging Deep Learning U-Net ResNet

3D Speckle Volume Quantifier

Custom analysis tool developed for my undergraduate honors thesis at Brown University in the Larschan Lab. Quantifies nuclear speckle volumes from confocal z-stacks to study how the GA-binding transcription factor CLAMP influences splicing condensate formation in Drosophila embryos.

Image Analysis Confocal Microscopy Python Biology

Mama's Pastaria (VR Game)

Final project for Columbia's COMS W4172: 3D User Interfaces and Augmented Reality. A Unity-based VR cooking simulator where players prepare pasta dishes for customers using interactive mechanics inspired by Papa's Pizzeria and Cooking Mama.

VR Unity C# Game Development

Desktop World

Final project for Brown's CSCI 1230: Computer Graphics. A procedurally generated snow globe–like world featuring terrain, particle systems, boids, custom shaders, and selective bloom effects.

Live demo: desktop-world.netlify.app

Computer Graphics WebGL Shaders Procedural Generation

WordleTerminal

A terminal-based clone of NYT Wordle, built for fun in Python.

Python Terminal Game