Biology subjects

Sosnick, L.

Publications and source records attributed to Sosnick, L..

3 recordsLinked to original sources

bioRxiv2026

Whole-genome 3D architectural screen reveals modulators of brain DNA structure

Three-dimensional (3D) genome architecture is the foundation of gene regulation, and plays a critical role in normal physiology and disease. However, our understanding of its biochemical determinants has long been limited by technology: imaging-based screens only profile a small number of loci, while sequencing-based studies rarely exceed 100 samples or conditions. Here we present "in-plate chromosome conformation capture" (Plate-C), a high-throughput, cost-effective platform that profiles thousands of whole-genome architectures in a day. Plate-C enabled the first chemical screen for whole-genome structural changes--profiling 2,956 samples from 834 conditions across 5 neuronal and glial types, accompanied by 6,081 single cells using "easy diploid chromosome conformation capture" (Easy Dip-C) and 200,893 single-cell transcriptomes. We discovered that diverse, dose/time-dependent, and cell type/species-specific modes of DNA structural changes can be rapidly induced by manipulating epigenetic (HDAC, BET), metabolic (mTOR), proteostatic (UPR), developmental (GSK3/Wnt, Hedgehog), immune (cGAS/STING), and neurotransmission pathways. To validate our finding in vivo, we demonstrated in newborn mice that HDAC inhibition drives brain-wide genome rewiring within hours, highly correlated with changes in vitro and inducing a latent structural and transcriptional state orthogonal to normal differentiation. By enabling massively parallel profiling of whole-genome structures, Plate-C paves the way for systematic discovery of DNA folding principles to better understand and engineer the human genome in 3D.

genomics↗

bioRxiv2026

BioReason-Pro: Advancing Protein Function Prediction with Multimodal Biological Reasoning

Protein function annotation is fundamental to understanding biological mechanisms, designing therapeutics, and advancing biomedical research. Current computational methods either rely on shallow sequence similarity or treat function prediction as isolated classification tasks, failing to capture the integrative reasoning across sequence, structure, domains, and interactions that expert biologists perform to infer function. We introduce BioReason-Pro, the first multimodal reasoning large language model (LLM) for protein function prediction that integrates protein embeddings with biological context to generate structured reasoning traces. A key input into BioReason-Pro is the set of GO term predictions made by GO-GPT, our autoregressive transformer that captures hierarchical and cross-aspect dependencies of GO terms. BioReason-Pro is trained via supervised fine-tuning on synthetic reasoning traces generated by GPT-5 for over 130K proteins and further optimized through reinforcement learning. It achieves 73.6% Fmax on GO term prediction and an LLM judge score of 8/10 on functional summaries, substantially outperforming previous methods. Evaluations with human protein experts show that BioReason-Pro annotations are preferred over ground truth UniProt annotations in 79% of cases. Remarkably, BioReason-Pro predicted a novel interaction partner for the renal cancer biomarker RCDG1, which we confirmed in the lab by co-immunoprecipitation. In other binding-partner predictions, its per-residue attention localized to the exact contact residues resolved in cryo-EM structures. Together, GO-GPT and BioReason-Pro establish a framework for protein function prediction that combines precise ontology modeling with interpretable biological reasoning.

molecular biology↗

bioRxiv2025

CRISPR-Cas13d-mediated targeting of a context-specific essential gene enables selective elimination of uveal melanoma

Uveal melanoma, the most common eye cancer in adults, remains limited to surgical intervention and chemotherapy, with a dismal survival rate that has not improved in over 50 years. To address this therapeutic impasse, we systematically analyzed public gene expression, RNAi, and CRISPR knockout datasets and identified RASGRP3 as an essential gene specifically for uveal melanoma. RasGRP3 is uniquely overexpressed and essential for survival in uveal melanoma cells, but dispensable in healthy cells. RasGRP3 remains "undruggable" due to its intracellular localization and lack of targetable binding pockets. To overcome this, we developed a CRISPR-Cas13d RNA-targeting therapeutic that specifically knocks down RasGRP3 mRNA. This Cas13d-based therapeutic mediates selective uveal melanoma killing through two synergistic mechanisms: (i) direct silencing of the essential RasGRP3 transcript, and (ii) collateral RNA degradation triggered by the cleavage of overexpressed RasGRP3. When delivered via optimized lipid nanoparticles encoding Cas13d mRNA and guide RNA, this strategy eliminated >97% of uveal melanoma cells while sparing healthy cells, including retinal pigment epithelial cells. This approach outperformed conventional Cas9 and siRNA methods in potency without inducing permanent genomic alterations. Our findings establish a RNA-targeting therapeutic for uveal melanoma and a framework for Cas13d-based interventions against broad "undruggable" cancers.

synthetic biology↗