perturbation response prediction

SpectroVQ improves proteomics data compression and peptide identification by using a noise-aware vector-quantized autoencoder to retain useful signals and enhance library searches.

This paper introduces a refined CRISPR screening protocol to compare cellular responses across multiple stressors, reducing false positives by 40% through fluorescence-based reporters and fixation steps.

This study identifies six microglial transcriptional profiles linked to specific morphological phenotypes and shows AD risk genes disrupt morpho-transcriptomic coupling and phagocytic capacity.

This study shows that silencing racR in E. coli triggers stress responses linked to prophage excision, revealing mechanisms of survival and cell death under lethal conditions.

A conditional latent diffusion transformer predicts immunofluorescence images of 12,800 human proteins from nucleus, ER and microtubule stains, outperforming PUPS and CELL-Diff, including on an unseen drug-treated cell line.

The paper benchmarks 13 models for predicting single-cell perturbation responses across 25 datasets, finding performance varies with perturbation effect size and evaluation metric.

GenoME predicts genomic profiles and perturbations using ATAC-seq and DNA sequence, generalizing to new cell types and identifying enhancer-promoter connections.

The paper shows co-inhibition of AGC kinases enhances CNS neuron regeneration and identifies TMP-316 as a drug candidate achieving this in a rat model.

This study shows that combining ACLY and OGDH inhibitors VPA and BA synergistically inhibits HCC cell lines with minimal toxicity to normal liver cells.

This study uses deep reinforcement learning to design treatment protocols for NSCLC that improve robustness, quality of life, and time to progression compared to standard adaptive therapy and maximum tolerable dose approaches.

This study uses CRISPRi Perturb-seq to analyze 1,408 autism risk genes in human cortical neurons, identifying 215 genes that cause global transcriptomic dysregulation and uncovering shared transcriptional pathways.

A set-aware latent transport model predicts treated cell populations from unpaired controls, keeps perturbation-specific signal in CRISPR, drug and cytokine data, and guided a cytokine choice tested in three donors.

This study optimizes CRISPR/Cas9 mutagenesis in Drosophila da neurons by developing uCas9 transgenes that reduce cytotoxicity while maintaining gene editing efficiency.

AnnFlux models single-cell perturbation dynamics using object-conditioned neural SDEs, predicting responses to unseen perturbations and combinations.

scDRP estimates individual treatment effects and predicts single-cell responses to perturbations using disentangled representation learning with a sparsity regularized β-VAE.

The paper develops uniSynK, a single-vector chemogenetic tool that optimizes the expression ratio of two components to robustly induce dendritic spine enlargement with fewer off-target effects.

State, a set-based transformer trained on over 100 million perturbed cells, predicts perturbation responses in new cellular contexts better than mean, linear and deep learning baselines.

DIPTAR converts protein degradation into a selectable phenotype to enable functional studies of degradation pathways and host-pathogen interactions.

This paper introduces a benchmark to evaluate single-cell drug perturbation models by assessing their fidelity to drug-response signatures, finding that expression similarity is a poor proxy for mechanism preservation.

The paper identifies a co-dependency between CDK12 and the BAF chromatin-remodeling complex in TNBC, showing that their co-inhibition synergistically reduces cancer cell viability.

MultiFlow predicts single-cell multiomic perturbation responses by jointly modeling gene expression and chromatin accessibility across cellular contexts.

This study shows INPP4B promotes gemcitabine resistance in PDAC by enhancing lysosomal exocytosis, which can be reversed by lysosome-targeting drugs.

This study evaluates scRNA-seq foundation models for plant perturbation response prediction, finding that performance varies significantly with evaluation protocol and that preserving gene identity improves generalization.

The paper develops a scarless conditional sgRNA system using endogenous RNA processing to enable precise temporal control of genome editing with improved efficiency and fidelity.

RegFM improves gene regulation modeling by integrating cis-regulatory elements and trans-acting regulators in human cells, outperforming existing methods in multiple tasks.

The paper screens bacterial cutinases to identify those with PET hydrolysis activity, finding it rare and mostly within the canonical-fold family.

The paper introduces MASCOT, a multi-agent framework for molecular optimization, which improves anesthetic candidates with better safety and recovery properties.

This paper introduces a self-regulating CRISPRi system that reduces leakiness while maintaining inducible gene repression, enabling portable and multiplexed applications.

VIPerturb-seq enables genome-wide CRISPR perturbation screens using probe-based detection, improving throughput and efficiency for single-cell analysis.

This paper introduces CIFM, a generative virtual tissue model that simulates transcriptional dynamics under therapeutic conditions to design immunotherapy strategies for cancer and autoimmune diseases.