AI x Bio Discovery

The FERM Guild: A Differentially Correlated Microbial Module Drives Hypertension via Metabolic Flux Perturbations

Figure 1

Figure 2

Figure 3

The FERM Guild: A Differentially Correlated Microbial Module Drives Hypertension via Metabolic Flux Perturbations [new]
Hypertension linked to gut microbiota; FERM guild drives it via metabolite changes.

Haruka Resolves Perturbation Response Heterogeneity in Spatial Cell Niches

Figure 1

Figure 2

Figure 3

Haruka Resolves Perturbation Response Heterogeneity in Spatial Cell Niches [new]
Identifies condition-specific vs. shared spatial domains in tissues using contrastive learning with spatial context.

Figure 1

Figure 2

Figure 4

Figure 3

EVRCEPT: EV RNA Cargo Enrichment Prediction Tool to predict enrichment of RNA into Extracellular Vesicles [new]
Predicts EV RNA enrichment via sequence features/RBP motifs. Finds RNA-RBP interactions (linear/circRNAs).

EMCF ecosystem: Towards pretrained foundation model for electron microscopy image analysis

Figure 1

Figure 2

Figure 3

EMCF ecosystem: Towards pretrained foundation model for electron microscopy image analysis [new]
EMCF tackles vEM bottlenecks using datasets, restoration, & analysis models. Improves nanoscale image quality & visualization.

IsoNet2 determines cellular structures at submolecular resolution without averaging

Figure 1

Figure 2

IsoNet2 determines cellular structures at submolecular resolution without averaging [new]
3D cryo-ET recon: self-sup. DL refines struct., revealing details in situ.

Design and experimental characterization of specificity-switching mutational paths of WW domains

Figure 1

Figure 2

Figure 3

Design and experimental characterization of specificity-switching mutational paths of WW domains [new]
WW domain specificity shifts via designed mutations. Binding assays confirm epistatic interactions are key.

Figure 1

Figure 2

Figure 3

Figure 4

AI recognizes convergent somatic hypermutation signatures to allow the discovery of variant-resilient broadly neutralizing antibodies [new]

Structure-based design of antibody repertoires with drug-like properties

Figure 1

Figure 2

Figure 3

Structure-based design of antibody repertoires with drug-like properties [new]
Structure-based design creates stable, drug-like antibody repertoires, a non-animal immunization alternative.

Graph Neural Networks Model Based on Atomic Hybridization for Predicting Drug Targets

Figure 1

Figure 3

Figure 2

Graph Neural Networks Model Based on Atomic Hybridization for Predicting Drug Targets [updated]
Models drug-target affinity by combining graph neural networks on atomic features with molecular descriptors for enhanced prediction.

A transcription factor-responsive enhancer discovery platform for targeted immunotherapy

Figure 2

Figure 3

Figure 4

A transcription factor-responsive enhancer discovery platform for targeted immunotherapy [new]
Cell-type enhancers ID'd for targeted immunotherapy & precise gene regulation.

MoDaH achieves rate optimal batch correction

Figure 2

Figure 3

Figure 4

MoDaH achieves rate optimal batch correction [new]
Corrects batch effects in single-cell data using a Gaussian mixture model. It guarantees minimax optimal error rates.

PhaLP 2.0: extending the community-oriented phage lysin database with a SUBLYME pipeline for metagenomic discovery

Figure 1

Figure 2

Figure 3

PhaLP 2.0: extending the community-oriented phage lysin database with a SUBLYME pipeline for metagenomic discovery [new]
PhaLP 2.0: SUBLYME pipeline integrates new virome lysin sequences, improving discovery.

SPARK: deciphering tumor-specific signaling networks through an integrative predictive model

Figure 1

Figure 2

Figure 3

SPARK: deciphering tumor-specific signaling networks through an integrative predictive model [new]
SPARK: multi-modal data predicts kinase-substrate links, revealing cancer signaling & therapeutic targets.

COFFEE-PRESC: A fast pre-screening method using compound retrieval by pairwise positional relationship of representative fragments

Figure 1

Figure 2

Figure 3

COFFEE-PRESC: A fast pre-screening method using compound retrieval by pairwise positional relationship of representative fragments [new]

GDA-Pred: Generative AI-Driven Data Augmentation for Improved Prediction of IL-6 and IL-13 Inducing Peptides

Figure 1

Figure 2

Figure 3

GDA-Pred: Generative AI-Driven Data Augmentation for Improved Prediction of IL-6 and IL-13 Inducing Peptides [new]
GDA-Pred: Enhances IL-6/13 prediction; uses GANs, DMs, & VAEs to augment data.

Probabilistic Modelling of Prime Editing Variant CorrectionEfficiency

Figure 1

Figure 2

Figure 3

Probabilistic Modelling of Prime Editing Variant CorrectionEfficiency [new]
Predicts prime editing efficiency with uncertainty, modeling outcomes in 3D simplex space & identifying key sequence features affecting results.

zifalsnm: Zero-Inflated Bayesian factor analysis model with skew-normal priors for modeling microbiome data

Figure 1

Figure 2

Figure 4

zifalsnm: Zero-Inflated Bayesian factor analysis model with skew-normal priors for modeling microbiome data [new]
Microbiome: Zero-inflated FA, dimension reduction, skew-normal priors 4 asymmetry.

Figure 1

Figure 2

Figure 3

Table 1

GUANinE v1.1 Reveals Complementarity of Supervised and Genomic Language Models [new]
Reveals that supervised and unsupervised genomic models excel at distinct tasks. Hybrid approaches may be the future.

Yomix: An Interactive Tool for the Exploration of Low-Dimensional Embeddings in Omics Data

Figure 1

Yomix: An Interactive Tool for the Exploration of Low-Dimensional Embeddings in Omics Data [new]
Explores omics embeddings via interactive visualization. Enables subset definition, signature computation, and distribution comparison.

WSInsight as a cloud-native pipeline for single-cell pathology inference on whole-slide images

Figure 1

Figure 2

Table 1

WSInsight as a cloud-native pipeline for single-cell pathology inference on whole-slide images [new]
Scalable, single-cell pathology inference via cloud pipeline w/ QC.

AlphaFold3 and Intrinsically Disordered Proteins: Reliable Monomer Prediction, Unpredictable Multimer Performance

Figure 1

Figure 2

Figure 4

AlphaFold3 and Intrinsically Disordered Proteins: Reliable Monomer Prediction, Unpredictable Multimer Performance [new]
AF3 predicts monomer disorder like AF2 (data-driven). Multimer performance varies due to priors.

Deep Learning-Based Drug Repurposing Using Knowledge Graph Embeddings and GraphRAG

Table I

Table II

Table III

Deep Learning-Based Drug Repurposing Using Knowledge Graph Embeddings and GraphRAG [new]
GraphRAG predicts & explains drug-disease links for repurposing using KG embeddings.

mmContext: an open framework for multimodal contrastive learning of omics and text data

Figure 1

mmContext: an open framework for multimodal contrastive learning of omics and text data [new]
mmContext: Framework uses contrastive learning to integrate omics/text data, w/ Sentence Transformers & HF.

scProtoTransformer: Scalable Reference Mapping Across Molecules, Cells and Donors

Figure 1

Figure 2

Figure 3

scProtoTransformer: Scalable Reference Mapping Across Molecules, Cells and Donors [new]
Uses a prototype-based Transformer to map single-cell data across resolutions. It projects gene expression into pathway prototypes.

Subtype-Specific Dependencies and Drug Vulnerabilities Enable Precision Therapeutics in Head and Neck Cancer

Figure 2

Figure 3

Figure 4

Subtype-Specific Dependencies and Drug Vulnerabilities Enable Precision Therapeutics in Head and Neck Cancer [new]
HNSCC subtypes: survival circuits ID'd, vulnerabilities mapped 4 targeted Rx.

Integration of single-cell transcriptome and genetic profiles reveals critical cell types and genes in inflammatory bowel disease

Figure 1

Figure 2

Figure 3

Integration of single-cell transcriptome and genetic profiles reveals critical cell types and genes in inflammatory bowel disease [new]

Figure 1

Figure 2

Figure 3

Figure 4

Protein Diffusion Models as Statistical Potentials [new]
Models protein conformational space using an energy-based approach. This allows for ranking structures, predicting conformations, and modeling mutations.

Figure 1

Figure 2

Figure 3

Figure 4

Spotsweeper-py: spatially-aware quality control metrics for spatial omics data in the Python ecosystem [new]
Spatial QC in Python: Locates artifacts, preserves tissue architecture.

The eXplainable Artificial Intelligence (XAI) Triad: Models, Importances, and Significance at Scale

Figure 1

Figure 2

Figure 3

The eXplainable Artificial Intelligence (XAI) Triad: Models, Importances, and Significance at Scale [new]
XAI triad: AI feature selection benchmarks model/importance/significance on simulated bio data.

Unimeth: A unified transformer framework for accurate DNA methylation detection from nanopore reads

Figure 1

Figure 2

Figure 3

Unimeth: A unified transformer framework for accurate DNA methylation detection from nanopore reads [new]
Transformer predicts DNA methylation from nanopore reads via multi-phase training for accuracy.