AI x Bio Discovery

Integrating morphology and gene expression of neural cells in unpaired single-cell data using GeoAdvAE

Figure 1

Figure 2

Figure 3

Integrating morphology and gene expression of neural cells in unpaired single-cell data using GeoAdvAE [updated]
aligns cell form & function in shared latent space, uncovers transcr. shifts & gene markers corr. w/ morph changes.

PhosSight: a Unified Deep Learning Framework Boosting and Accelerating Phosphoproteome Identification to Enable Biological Discoveries

Figure 1

Figure 2

Figure 3

PhosSight: a Unified Deep Learning Framework Boosting and Accelerating Phosphoproteome Identification to Enable Biological Discoveries [new]

Joint Geometric-Chemical Distance for Protein Surfaces

Figure 1

Figure 2

Table I

Joint Geometric-Chemical Distance for Protein Surfaces [new]
...integrates structural and physicochemical discrepancies via surface alignment of geometry and chemical fields, unifying functional interface comparison.

Reaction-Conditioned Enzyme Discovery with Multimodal Deep Learning

Figure 1

Figure 2

Figure 3

Reaction-Conditioned Enzyme Discovery with Multimodal Deep Learning [new]
Unifies reaction encoding w/ protein language models to enable zero-shot discovery of enzymes for previously unseen chemical reactions, moving beyond homology.

torch-projectors: A High-Performance Differentiable Projection Library for PyTorch

Table 1

Table 2

Table 3

torch-projectors: A High-Performance Differentiable Projection Library for PyTorch [new]
provides differentiable Fourier-space projections for EM (single-particle analysis, electron tomography) with 2D/3D operators and grad. support.

Developing SCL2205 : A Protein Sequence-based Spatial Modelling Dataset for the Protein Language Model Frontier [new]
High-quality dataset for robust DL protein localization modeling, stringent partitioned to minimize leakage.

Intrinsic dataset features drive mutational effect prediction by protein language models

Figure 1

Figure 2

Figure 3

Intrinsic dataset features drive mutational effect prediction by protein language models [new]
via site var. metrics explaining dataset diffs (e.g., viral/cellular) & reveal models leverage site-specific effects, not broad patterns.

Inferring large networks with matrix factorisation to capture non-linear dependencies among genes using sparse single-cell profiles

Figure 1

Figure 2

Figure 3

Inferring large networks with matrix factorisation to capture non-linear dependencies among genes using sparse single-cell profiles [new]

Generating Hybrid Proteins with the MSA-Transformer

Figure 1

Figure 2

Figure 4

Generating Hybrid Proteins with the MSA-Transformer [updated]
...via an iterative framework that creates intermediate sequences between homologous pairs, integrating properties and exploring novel structural permutations.

Figure 2

Figure 3

Figure 4

Figure 5

A Universal, AI-based Design Framework for Efficient Manufacturing of mRNA Therapeutics [new]
enables UD predicting mRNA manufacturability, learning molecular mechanisms to accel. and democratize therapeutic development.

Quantifying Memorization and Privacy Risks in Genomic Language Models

Figure 1

Figure 2

Figure 3

Quantifying Memorization and Privacy Risks in Genomic Language Models [new]
...is assessed by a multi-vector framework combining perplexity, canary extraction, and membership inference to audit GLMs for sequence memorization.

Sampling on Discrete Spaces with Temporal Point Processes [new]
Sampling on Discrete Spaces with Temporal Point Processes, structured as coupled infinite-server queues, whose event counts converge to target distributions.

FAMUS: A Few-Shot Learning Framework for Large-Scale Protein Annotation [new]
...transforms HMM similarity scores via contrastive learning into a condensed vector space, assigning protein functions by leveraging all profile hits.

Abstract 1

Figure 2

Figure 3

Table 1

Improving Causal Gene Identification Using Large Language Models [new]
...by integrating retrieval-augmented generation and genomic distance features to better pinpoint causal genes within GWAS-identified loci.

Bacterial proteome foundation model enhances functional prediction from enzymes to ecological interactions

Figure 1

Figure 2

Figure 3

Bacterial proteome foundation model enhances functional prediction from enzymes to ecological interactions [new]
..., by learning ctx gene embs & org-level reps from 10k genomes, revealing gene interactions & functional landscapes.

From General-Purpose to Disease-Specific Features: Aligning LLM Embeddings on a Disease-Specific Biomedical Knowledge Graph for Drug Repurposing

Table 1

Figure 1

Figure 2

From General-Purpose to Disease-Specific Features: Aligning LLM Embeddings on a Disease-Specific Biomedical Knowledge Graph for Drug Repurposing [new]

Figure 2

Figure 3

Figure 4

Figure 5

Counting strands in outer membrane beta-barrels [new]
by refining an algorithm integrating vector angles, H-bonds, & connectivity, enabling large-scale analysis of their distribution and evolution.

NeuroNarrator: A Generalist EEG-to-Text Foundation Model for Clinical Interpretation via Spectro-Spatial Grounding and Temporal State-Space Reasoning

Figure 2

Figure 3

Figure 5

NeuroNarrator: A Generalist EEG-to-Text Foundation Model for Clinical Interpretation via Spectro-Spatial Grounding and Temporal State-Space Reasoning [new]

Neurotox: Deep learning decodes conserved hallmarks of neurotoxicity across venomous species

Figure 1

Figure 2

Neurotox: Deep learning decodes conserved hallmarks of neurotoxicity across venomous species [new]
Neurotoxicity arises from distributed sequence features shaping 2ndary struct. & receptor interact., not isolated contact residues.

Integrated proteomic screening reveals design principles of CRBN molecular glue degraders

Figure 1

Figure 3

Figure 2

Integrated proteomic screening reveals design principles of CRBN molecular glue degraders [new]
Identified 124 new CRBN neosubstrates, many non-G-loop, mapping chem structs to target selectivity, framework for rational MGD design.

InversePep: Diffusion-Driven Structure-Based Inverse Folding for Functional Peptides

Figure 1

Figure 2

Figure 3

InversePep: Diffusion-Driven Structure-Based Inverse Folding for Functional Peptides [new]
...generates diverse sequences that reliably adopt target peptide 3D backbone conformations.

MOZAIC: Compound Growth via In Silico Reactions and Global Optimization using Conformational Space Annealing

Figure 1

Figure 2

Figure 3

MOZAIC: Compound Growth via In Silico Reactions and Global Optimization using Conformational Space Annealing [new]
...for FBDD, producing synth., diverse mols with balanced lead-like properties overcoming synthetic pathway issues.

Physics-informed multi-encoder adaptive optics enables rapid aberration correction for intravital microscopy of deep complex tissue

Figure 1

Figure 2

Figure 3

Physics-informed multi-encoder adaptive optics enables rapid aberration correction for intravital microscopy of deep complex tissue [new]

SpatioCAD: Context-aware graph diffusion model for pinpointing spatially variable genes in heterogeneous tissues

Figure 1

Figure 2

Figure 3

SpatioCAD: Context-aware graph diffusion model for pinpointing spatially variable genes in heterogeneous tissues [new]
...by decoupling genuine spatial expression from confounding cell density variations using graph diffusion model.

Fung-AI: An AI/ML-driven pipeline for antifungal peptide discovery

Figure 1

Figure 2

Figure 3

Fung-AI: An AI/ML-driven pipeline for antifungal peptide discovery [new]
...deploys a GAN to generate novel antifungal peptides, then filters them with in silico classifiers for experimental validation.

STAR Suite: Integrating transcriptomics through AI software engineering in the NIH MorPhiC consortium

Figure 1

Figure 2

Figure 3

STAR Suite: Integrating transcriptomics through AI software engineering in the NIH MorPhiC consortium [new]
... by modernizing its core C++ codebase to unify diverse functionalities, addressing data scalability challenges.

Computed atlas of the human GPCR-G protein signaling complexes

Figure 1

Figure 2

Figure 3

Computed atlas of the human GPCR-G protein signaling complexes [new]
This 3D atlas predicts GPCR-G-protein coupling specificity across human GPCRome, revealing mechanisms for uncharacterized receptors and tissue-specific signaling.

From games to biology and beyond: 10 years of AlphaGo’s impact](blog]
... is catalyzing scientific discovery and paving the path to artificial general intelligence.

SwiftTCR: Efficient Computational Docking protocol of TCRpMHC-I Complexes Using Restricted Rotation Matrices

Figure 1

Figure 2

Figure 3

SwiftTCR: Efficient Computational Docking protocol of TCRpMHC-I Complexes Using Restricted Rotation Matrices [updated]
Uses consistent TCR docking angles to rapidly generate structural models for T-cell recog. studies and therapy design.

screenshot

Population-level structural variant characterization using pangenome graphs [new]
...employs a recurrent neural network to identify diverse, complex SV patterns directly from these graphs.