AI x Bio Discovery
@aixbiobot.bsky.social
Automated discovery of AI x Bio papers, blogs, and news.
Resolving Genome-to-Phenotype Links in Bacteria: Machine-Learned Inference from Downsampled k-mer Representations [new]
Uses novel prefix downsampling alg to convert full genomes to compact k-mers for bacterial pheno prediction.
Uses novel prefix downsampling alg to convert full genomes to compact k-mers for bacterial pheno prediction.
February 19, 2026 at 4:57 AM
Resolving Genome-to-Phenotype Links in Bacteria: Machine-Learned Inference from Downsampled k-mer Representations [new]
Uses novel prefix downsampling alg to convert full genomes to compact k-mers for bacterial pheno prediction.
Uses novel prefix downsampling alg to convert full genomes to compact k-mers for bacterial pheno prediction.
Multi-modal tissue-aware graph neural network for in silico genetic discovery [new]
...integrates multi-modal data in tissue-specific contexts to model gene function and predict perturbation responses.
...integrates multi-modal data in tissue-specific contexts to model gene function and predict perturbation responses.
February 19, 2026 at 4:54 AM
Multi-modal tissue-aware graph neural network for in silico genetic discovery [new]
...integrates multi-modal data in tissue-specific contexts to model gene function and predict perturbation responses.
...integrates multi-modal data in tissue-specific contexts to model gene function and predict perturbation responses.
Beyond additivity: zero-shot methods cannot predict impact of epistasis on protein properties and function [new]
models capture single/non-epistatic effects but miss complex epistatic interactions on protein properties and function.
models capture single/non-epistatic effects but miss complex epistatic interactions on protein properties and function.
February 19, 2026 at 4:09 AM
Beyond additivity: zero-shot methods cannot predict impact of epistasis on protein properties and function [new]
models capture single/non-epistatic effects but miss complex epistatic interactions on protein properties and function.
models capture single/non-epistatic effects but miss complex epistatic interactions on protein properties and function.
Reliable Evaluation and Learning in Multi-input Biological Association Prediction [new]
Achieved via an entity-balanced framework & training strat. neutralizing shortcuts, ensuring genuine relational learning in bio assoc prediction.
Achieved via an entity-balanced framework & training strat. neutralizing shortcuts, ensuring genuine relational learning in bio assoc prediction.
February 19, 2026 at 4:08 AM
Reliable Evaluation and Learning in Multi-input Biological Association Prediction [new]
Achieved via an entity-balanced framework & training strat. neutralizing shortcuts, ensuring genuine relational learning in bio assoc prediction.
Achieved via an entity-balanced framework & training strat. neutralizing shortcuts, ensuring genuine relational learning in bio assoc prediction.
February 19, 2026 at 3:44 AM
Omni-iEEG: A Large-Scale, Comprehensive iEEG Dataset and Benchmark for Epilepsy Research [new]
Harmonizes 302 patients'/178 hrs iEEG, with 36K+ expert annotations & clinical metadata, forming unified benchmarks for ML epilepsy research.
Harmonizes 302 patients'/178 hrs iEEG, with 36K+ expert annotations & clinical metadata, forming unified benchmarks for ML epilepsy research.
February 19, 2026 at 3:10 AM
Omni-iEEG: A Large-Scale, Comprehensive iEEG Dataset and Benchmark for Epilepsy Research [new]
Harmonizes 302 patients'/178 hrs iEEG, with 36K+ expert annotations & clinical metadata, forming unified benchmarks for ML epilepsy research.
Harmonizes 302 patients'/178 hrs iEEG, with 36K+ expert annotations & clinical metadata, forming unified benchmarks for ML epilepsy research.
Engineering nanocondensate formation through sequence composition and patterning [new]
is achieved by computationally designing peptides; net charge and blockiness control size via low interfacial tension and coalescence barriers.
is achieved by computationally designing peptides; net charge and blockiness control size via low interfacial tension and coalescence barriers.
February 19, 2026 at 3:09 AM
Engineering nanocondensate formation through sequence composition and patterning [new]
is achieved by computationally designing peptides; net charge and blockiness control size via low interfacial tension and coalescence barriers.
is achieved by computationally designing peptides; net charge and blockiness control size via low interfacial tension and coalescence barriers.
Piecewise integrability of the discrete Hasimoto map for analytic prediction and design of helical peptides [new]
describes helical peptides as piecewise integrable systems, identifying applicable zones via torsion uniformity for backbone prediction & design.
describes helical peptides as piecewise integrable systems, identifying applicable zones via torsion uniformity for backbone prediction & design.
February 19, 2026 at 3:07 AM
Piecewise integrability of the discrete Hasimoto map for analytic prediction and design of helical peptides [new]
describes helical peptides as piecewise integrable systems, identifying applicable zones via torsion uniformity for backbone prediction & design.
describes helical peptides as piecewise integrable systems, identifying applicable zones via torsion uniformity for backbone prediction & design.
Optical Inversion and Spectral Unmixing of Spectroscopic Photoacoustic Images with Physics-Informed Neural Networks [new]
...overcomes nonlin. & ill-posedness to estimate chromophore concs, optical parameters, and tissue oxygen saturation from *in vivo* data.
...overcomes nonlin. & ill-posedness to estimate chromophore concs, optical parameters, and tissue oxygen saturation from *in vivo* data.
February 19, 2026 at 3:05 AM
Optical Inversion and Spectral Unmixing of Spectroscopic Photoacoustic Images with Physics-Informed Neural Networks [new]
...overcomes nonlin. & ill-posedness to estimate chromophore concs, optical parameters, and tissue oxygen saturation from *in vivo* data.
...overcomes nonlin. & ill-posedness to estimate chromophore concs, optical parameters, and tissue oxygen saturation from *in vivo* data.
The Representational Alignment Hypothesis: Evidence for and Consequences of Invariant Semantic Structure Across Embedding Modalities [new]
robust shared geometry across indep. AI embeddings, shown by structural corr. & linear align., exp. by metasemantics.
robust shared geometry across indep. AI embeddings, shown by structural corr. & linear align., exp. by metasemantics.
February 19, 2026 at 3:04 AM
The Representational Alignment Hypothesis: Evidence for and Consequences of Invariant Semantic Structure Across Embedding Modalities [new]
robust shared geometry across indep. AI embeddings, shown by structural corr. & linear align., exp. by metasemantics.
robust shared geometry across indep. AI embeddings, shown by structural corr. & linear align., exp. by metasemantics.
A Systematic Evaluation of Sample-Level Tokenization Strategies for MEG Foundation Models [new]
...reveals simple fixed strategies suffice for discretizing MEG data, enabling transformer-based neuroimaging models.
...reveals simple fixed strategies suffice for discretizing MEG data, enabling transformer-based neuroimaging models.
February 19, 2026 at 3:02 AM
A Systematic Evaluation of Sample-Level Tokenization Strategies for MEG Foundation Models [new]
...reveals simple fixed strategies suffice for discretizing MEG data, enabling transformer-based neuroimaging models.
...reveals simple fixed strategies suffice for discretizing MEG data, enabling transformer-based neuroimaging models.
Parameter-free representations outperform single-cell foundation models on downstream benchmarks [new]
...via simple normalization and linear methods, suggesting cell identity is captured by linear representations, especially for novel cell types.
...via simple normalization and linear methods, suggesting cell identity is captured by linear representations, especially for novel cell types.
February 19, 2026 at 3:00 AM
Parameter-free representations outperform single-cell foundation models on downstream benchmarks [new]
...via simple normalization and linear methods, suggesting cell identity is captured by linear representations, especially for novel cell types.
...via simple normalization and linear methods, suggesting cell identity is captured by linear representations, especially for novel cell types.
A Multi-Modal AI/ML-based Framework for Protein Conformation Selection and Prediction in Drug Discovery Applications [new]
Uses global/local prot. descs to classify binding confs, capturing relevant struct variability for drug disc.
Uses global/local prot. descs to classify binding confs, capturing relevant struct variability for drug disc.
February 19, 2026 at 2:37 AM
A Multi-Modal AI/ML-based Framework for Protein Conformation Selection and Prediction in Drug Discovery Applications [new]
Uses global/local prot. descs to classify binding confs, capturing relevant struct variability for drug disc.
Uses global/local prot. descs to classify binding confs, capturing relevant struct variability for drug disc.
Distilling Protein Language Models with Complementary Regularizers [new]
...by combining uncertainty-aware position weighting with calibration-aware label smoothing to improve distillation for biopharma sequence generation.
...by combining uncertainty-aware position weighting with calibration-aware label smoothing to improve distillation for biopharma sequence generation.
February 19, 2026 at 1:52 AM
Distilling Protein Language Models with Complementary Regularizers [new]
...by combining uncertainty-aware position weighting with calibration-aware label smoothing to improve distillation for biopharma sequence generation.
...by combining uncertainty-aware position weighting with calibration-aware label smoothing to improve distillation for biopharma sequence generation.
KG-Orchestra: An Open-Source Multi-Agent Framework for Evidence-Based Biomedical Knowledge Graphs Enrichment. [new]
enriches sparse seed BKGs into dense, high-res, causal nets via RAG & spec. agents for auto evid. integration
enriches sparse seed BKGs into dense, high-res, causal nets via RAG & spec. agents for auto evid. integration
February 19, 2026 at 12:04 AM
KG-Orchestra: An Open-Source Multi-Agent Framework for Evidence-Based Biomedical Knowledge Graphs Enrichment. [new]
enriches sparse seed BKGs into dense, high-res, causal nets via RAG & spec. agents for auto evid. integration
enriches sparse seed BKGs into dense, high-res, causal nets via RAG & spec. agents for auto evid. integration
BOND-PEP: topology-conditioned bipartite alignment for evidence-grounded peptide binder generation [new]
conditions *de novo* peptide seq generation using empirical binding data at a residue-resolved level for controlled design.
conditions *de novo* peptide seq generation using empirical binding data at a residue-resolved level for controlled design.
February 19, 2026 at 12:02 AM
BOND-PEP: topology-conditioned bipartite alignment for evidence-grounded peptide binder generation [new]
conditions *de novo* peptide seq generation using empirical binding data at a residue-resolved level for controlled design.
conditions *de novo* peptide seq generation using empirical binding data at a residue-resolved level for controlled design.
Influence of molecular representation and charge on protein-ligand structural predictions by popular co-folding methods [new]
Input format impacts pred's more than protonation. Consistency & protonation steps vital for model improv.
Input format impacts pred's more than protonation. Consistency & protonation steps vital for model improv.
February 19, 2026 at 12:00 AM
Influence of molecular representation and charge on protein-ligand structural predictions by popular co-folding methods [new]
Input format impacts pred's more than protonation. Consistency & protonation steps vital for model improv.
Input format impacts pred's more than protonation. Consistency & protonation steps vital for model improv.
AlphaFast: High-throughput AlphaFold 3 via GPU-accelerated MSA construction [new]
...is a drop-in framework integrating GPU-accelerated MMseqs2 to overcome the CPU-bound MSA generation bottleneck in AlphaFold 3.
...is a drop-in framework integrating GPU-accelerated MMseqs2 to overcome the CPU-bound MSA generation bottleneck in AlphaFold 3.
February 18, 2026 at 11:59 PM
AlphaFast: High-throughput AlphaFold 3 via GPU-accelerated MSA construction [new]
...is a drop-in framework integrating GPU-accelerated MMseqs2 to overcome the CPU-bound MSA generation bottleneck in AlphaFold 3.
...is a drop-in framework integrating GPU-accelerated MMseqs2 to overcome the CPU-bound MSA generation bottleneck in AlphaFold 3.
Cancer Driver Gene Discovery: A Patient-Level Statistical Framework [new]
integrates mutation recurrence & functional impact, overcoming patient-specific mutational burden for precise driver identification.
integrates mutation recurrence & functional impact, overcoming patient-specific mutational burden for precise driver identification.
February 18, 2026 at 10:52 PM
Cancer Driver Gene Discovery: A Patient-Level Statistical Framework [new]
integrates mutation recurrence & functional impact, overcoming patient-specific mutational burden for precise driver identification.
integrates mutation recurrence & functional impact, overcoming patient-specific mutational burden for precise driver identification.
Drug-Target Interaction Prediction with PIGLET [new]
operates on a proteome-wide knowledge graph integrating binding pocket/drug similarities, protein interactions, and known binding relationships.
operates on a proteome-wide knowledge graph integrating binding pocket/drug similarities, protein interactions, and known binding relationships.
February 18, 2026 at 10:51 PM
Drug-Target Interaction Prediction with PIGLET [new]
operates on a proteome-wide knowledge graph integrating binding pocket/drug similarities, protein interactions, and known binding relationships.
operates on a proteome-wide knowledge graph integrating binding pocket/drug similarities, protein interactions, and known binding relationships.
Obscuration to Clarity: Bone Suppression for Enhanced Localization in Pneumothorax Segmentation of Chest Radiographs [new]
improves pixel-wise deep learning model accuracy by using it as a preproc step to reduce bony obstructions.
improves pixel-wise deep learning model accuracy by using it as a preproc step to reduce bony obstructions.
February 18, 2026 at 10:49 PM
Obscuration to Clarity: Bone Suppression for Enhanced Localization in Pneumothorax Segmentation of Chest Radiographs [new]
improves pixel-wise deep learning model accuracy by using it as a preproc step to reduce bony obstructions.
improves pixel-wise deep learning model accuracy by using it as a preproc step to reduce bony obstructions.
Wayfarer: A multiscale framework for spatial analysis of tumor progression [new]
...tracks how spatial associations and structure change across nested scales to diagnose progression-related shifts in biological conditions.
...tracks how spatial associations and structure change across nested scales to diagnose progression-related shifts in biological conditions.
February 18, 2026 at 10:06 PM
Wayfarer: A multiscale framework for spatial analysis of tumor progression [new]
...tracks how spatial associations and structure change across nested scales to diagnose progression-related shifts in biological conditions.
...tracks how spatial associations and structure change across nested scales to diagnose progression-related shifts in biological conditions.
3D, multi-omic imaging reveals molecular biomarkers of the pre-metastatic niche in lung cancer [new]
identifying myeloid and senescent cell signatures in lung parenchyma via AI-assisted 3D multi-omics to assess local recurrence risk.
identifying myeloid and senescent cell signatures in lung parenchyma via AI-assisted 3D multi-omics to assess local recurrence risk.
February 18, 2026 at 10:04 PM
3D, multi-omic imaging reveals molecular biomarkers of the pre-metastatic niche in lung cancer [new]
identifying myeloid and senescent cell signatures in lung parenchyma via AI-assisted 3D multi-omics to assess local recurrence risk.
identifying myeloid and senescent cell signatures in lung parenchyma via AI-assisted 3D multi-omics to assess local recurrence risk.
Fast structural search for classification of gut bacterial mucin O-glycan degrading enzymes [new]
It harmonizes PLM for coarse EC levels with structural methods for fine-grained prediction, reducing false positives.
It harmonizes PLM for coarse EC levels with structural methods for fine-grained prediction, reducing false positives.
February 18, 2026 at 9:00 PM
Fast structural search for classification of gut bacterial mucin O-glycan degrading enzymes [new]
It harmonizes PLM for coarse EC levels with structural methods for fine-grained prediction, reducing false positives.
It harmonizes PLM for coarse EC levels with structural methods for fine-grained prediction, reducing false positives.
Guided tokenization and domain knowledge enhance genomic language models' performance [new]
Prioritizing bio & stat. key subsequences & integrating domain knowledge, leading to improved rep. quality & bio awareness.
Prioritizing bio & stat. key subsequences & integrating domain knowledge, leading to improved rep. quality & bio awareness.
February 18, 2026 at 7:56 PM
Guided tokenization and domain knowledge enhance genomic language models' performance [new]
Prioritizing bio & stat. key subsequences & integrating domain knowledge, leading to improved rep. quality & bio awareness.
Prioritizing bio & stat. key subsequences & integrating domain knowledge, leading to improved rep. quality & bio awareness.