Raktim Mitra

Chaperone, a DataFrame-based framework and agent harness for protein design tools, cut tokens 86% and GPU-hours 64% while raising pass rate from 6% to 12%.

This paper designs sequence-specific DNA binding proteins using deep learning, achieving a 100-fold improvement in success rates over previous methods.

This paper introduces NA-MPNN, a unified deep learning model for nucleic acid inverse folding, which outperforms existing methods in RNA sequence design and protein-DNA specificity prediction.

An all-atom diffusion model using a 14-atom residue representation and sparse U-Net architecture accelerates biomolecular design 10-fold, enabling experimentally validated DNA-binding proteins and multi-turnover cysteine hydrolases.

AtomWorks, a modular atom-level data framework, trains structure prediction and design models; the resulting RF3 improves chirality handling and performs between AF3 and Boltz.

DeepPBS predicts protein-DNA binding specificity using geometric deep learning, showing alignment with experimental mutagenesis data and enabling analysis of molecular dynamics simulations.

PNAbind predicts protein-nucleic acid binding using graph neural networks, achieving high accuracy in binding site prediction and discrimination between DNA and RNA specificity.

The paper introduces Tejaas, a method for detecting trans-eQTLs by reverse regression, identifying 18851 unique trans-eQTLs across 49 tissues and showing enrichment in regulatory regions.

RVAgene is a recurrent variational autoencoder that models gene expression time series data, accurately reconstructing temporal profiles and enabling biological feature discovery.