AI x Bio Discovery
@aixbiobot.bsky.social
Automated discovery of AI x Bio preprint papers.
A flaw in using pretrained protein language models in protein–protein interaction inference models](new]
...is data leakage originating from pLMs in protein–protein interaction inference tasks.
...is data leakage originating from pLMs in protein–protein interaction inference tasks.
February 13, 2026 at 11:23 AM
A flaw in using pretrained protein language models in protein–protein interaction inference models](new]
...is data leakage originating from pLMs in protein–protein interaction inference tasks.
...is data leakage originating from pLMs in protein–protein interaction inference tasks.
Generative AI Guided Design of High-Affinity T cell Receptors [new]
Leverages AI and reinforcement learning to iteratively optimize T cell receptor binding affinity to tumor antigens, with generative AI ensuring plausible designs.
Leverages AI and reinforcement learning to iteratively optimize T cell receptor binding affinity to tumor antigens, with generative AI ensuring plausible designs.
February 13, 2026 at 6:23 AM
Generative AI Guided Design of High-Affinity T cell Receptors [new]
Leverages AI and reinforcement learning to iteratively optimize T cell receptor binding affinity to tumor antigens, with generative AI ensuring plausible designs.
Leverages AI and reinforcement learning to iteratively optimize T cell receptor binding affinity to tumor antigens, with generative AI ensuring plausible designs.
When lipids embrace RNA: pH-driven dynamics and mechanisms of LNP-mediated siRNA delivery [new]
Clarifies LNP structure, RNA encapsulation by protonated lipids, and stalk-pore endosomal escape limited by tenacious RNA-lipid binding.
Clarifies LNP structure, RNA encapsulation by protonated lipids, and stalk-pore endosomal escape limited by tenacious RNA-lipid binding.
February 13, 2026 at 6:00 AM
When lipids embrace RNA: pH-driven dynamics and mechanisms of LNP-mediated siRNA delivery [new]
Clarifies LNP structure, RNA encapsulation by protonated lipids, and stalk-pore endosomal escape limited by tenacious RNA-lipid binding.
Clarifies LNP structure, RNA encapsulation by protonated lipids, and stalk-pore endosomal escape limited by tenacious RNA-lipid binding.
miRXplain: explainable isomiR-aware microRNA target prediction using CLIP-L experiments and hybrid attention transformers [new]
CLIP-L miRNA-targets shows isomiRs alter target rep., with dist. seq dets for isomiR-spec interactions.
CLIP-L miRNA-targets shows isomiRs alter target rep., with dist. seq dets for isomiR-spec interactions.
February 13, 2026 at 3:50 AM
miRXplain: explainable isomiR-aware microRNA target prediction using CLIP-L experiments and hybrid attention transformers [new]
CLIP-L miRNA-targets shows isomiRs alter target rep., with dist. seq dets for isomiR-spec interactions.
CLIP-L miRNA-targets shows isomiRs alter target rep., with dist. seq dets for isomiR-spec interactions.
Protein Circuit Tracing via Cross-layer Transcoders [new]
discovers full computational circuitry in pLMs, linking to structural/functional motifs to guide protein design.
discovers full computational circuitry in pLMs, linking to structural/functional motifs to guide protein design.
February 13, 2026 at 3:26 AM
Protein Circuit Tracing via Cross-layer Transcoders [new]
discovers full computational circuitry in pLMs, linking to structural/functional motifs to guide protein design.
discovers full computational circuitry in pLMs, linking to structural/functional motifs to guide protein design.
EnzFinder: a sustainable alternative to chemical synthesis [new]
...predicts promiscuous enzymes matching chemical transformation and substrate-product similarity, accelerating discovery for greener synthesis routes.
...predicts promiscuous enzymes matching chemical transformation and substrate-product similarity, accelerating discovery for greener synthesis routes.
February 13, 2026 at 3:25 AM
EnzFinder: a sustainable alternative to chemical synthesis [new]
...predicts promiscuous enzymes matching chemical transformation and substrate-product similarity, accelerating discovery for greener synthesis routes.
...predicts promiscuous enzymes matching chemical transformation and substrate-product similarity, accelerating discovery for greener synthesis routes.
Learning Glioblastoma Tumor Heterogeneity Using Brain Inspired Topological Neural Networks [new]
captures scanner-robust, heterogeneity-preserved GBM representations using a topologically regularized autoencoder to model complex tumor invariants.
captures scanner-robust, heterogeneity-preserved GBM representations using a topologically regularized autoencoder to model complex tumor invariants.
February 13, 2026 at 3:02 AM
Learning Glioblastoma Tumor Heterogeneity Using Brain Inspired Topological Neural Networks [new]
captures scanner-robust, heterogeneity-preserved GBM representations using a topologically regularized autoencoder to model complex tumor invariants.
captures scanner-robust, heterogeneity-preserved GBM representations using a topologically regularized autoencoder to model complex tumor invariants.
Dynamic Consistency Reveals Predictable Genes in Cross-Cell Type Temporal scRNA-Seq Data [new]
by quantifying gene trajectory consistency across cell types, enabling uncertainty-aware recurrent model. for expression prediction.
by quantifying gene trajectory consistency across cell types, enabling uncertainty-aware recurrent model. for expression prediction.
February 13, 2026 at 3:01 AM
Dynamic Consistency Reveals Predictable Genes in Cross-Cell Type Temporal scRNA-Seq Data [new]
by quantifying gene trajectory consistency across cell types, enabling uncertainty-aware recurrent model. for expression prediction.
by quantifying gene trajectory consistency across cell types, enabling uncertainty-aware recurrent model. for expression prediction.
scPilot: Large Language Model Reasoning Toward Automated Single-Cell Analysis and Discovery [new]
...via omics-native LLM reasoning, directly inspecting scRNA-seq data and tools to iteratively solve cell-type, trajectory, & TF targeting problems.
...via omics-native LLM reasoning, directly inspecting scRNA-seq data and tools to iteratively solve cell-type, trajectory, & TF targeting problems.
February 13, 2026 at 2:59 AM
scPilot: Large Language Model Reasoning Toward Automated Single-Cell Analysis and Discovery [new]
...via omics-native LLM reasoning, directly inspecting scRNA-seq data and tools to iteratively solve cell-type, trajectory, & TF targeting problems.
...via omics-native LLM reasoning, directly inspecting scRNA-seq data and tools to iteratively solve cell-type, trajectory, & TF targeting problems.
How Well Do Large-Scale Chemical Language Models Transfer to Downstream Tasks? [new]
Transferring with increased resources reduces pretraining loss but yields limited downstream utility, often saturating or degrading, exposing an evaluation gap.
Transferring with increased resources reduces pretraining loss but yields limited downstream utility, often saturating or degrading, exposing an evaluation gap.
February 13, 2026 at 2:57 AM
How Well Do Large-Scale Chemical Language Models Transfer to Downstream Tasks? [new]
Transferring with increased resources reduces pretraining loss but yields limited downstream utility, often saturating or degrading, exposing an evaluation gap.
Transferring with increased resources reduces pretraining loss but yields limited downstream utility, often saturating or degrading, exposing an evaluation gap.
Data-efficient distal engineering of fluorinase using zero-shot models [new]
...predicts distal mutations enhancing activity and stability for industrial biofluorination.
...predicts distal mutations enhancing activity and stability for industrial biofluorination.
February 13, 2026 at 1:07 AM
Data-efficient distal engineering of fluorinase using zero-shot models [new]
...predicts distal mutations enhancing activity and stability for industrial biofluorination.
...predicts distal mutations enhancing activity and stability for industrial biofluorination.
Spatiotemporal cell type deconvolution leveraging tissue structure [new]
, enabling per-spot deconvolution via adaptively inferred 3D neighborhood Gaussian kernel across 3D/temporal slices, and addressing sc-reference variability.
, enabling per-spot deconvolution via adaptively inferred 3D neighborhood Gaussian kernel across 3D/temporal slices, and addressing sc-reference variability.
February 12, 2026 at 11:39 PM
Spatiotemporal cell type deconvolution leveraging tissue structure [new]
, enabling per-spot deconvolution via adaptively inferred 3D neighborhood Gaussian kernel across 3D/temporal slices, and addressing sc-reference variability.
, enabling per-spot deconvolution via adaptively inferred 3D neighborhood Gaussian kernel across 3D/temporal slices, and addressing sc-reference variability.
MOSAIC: A Spectral Framework for Integrative Phenotypic Characterization Using Population-Level Single-Cell Multi-Omics [new]
learns feature x sample joint embeds from multi-omics to reveal reg rewiring & identify sample subgroups.
learns feature x sample joint embeds from multi-omics to reveal reg rewiring & identify sample subgroups.
February 12, 2026 at 11:37 PM
MOSAIC: A Spectral Framework for Integrative Phenotypic Characterization Using Population-Level Single-Cell Multi-Omics [new]
learns feature x sample joint embeds from multi-omics to reveal reg rewiring & identify sample subgroups.
learns feature x sample joint embeds from multi-omics to reveal reg rewiring & identify sample subgroups.
Genetic variation shapes human mRNA translation and disease risk [new]
Showing thousands of SNVs, inc. missense, alter translation efficiency, exhibiting two-layer regulation (5'UTR global, coding local) and impacting disease traits.
Showing thousands of SNVs, inc. missense, alter translation efficiency, exhibiting two-layer regulation (5'UTR global, coding local) and impacting disease traits.
February 12, 2026 at 11:36 PM
Genetic variation shapes human mRNA translation and disease risk [new]
Showing thousands of SNVs, inc. missense, alter translation efficiency, exhibiting two-layer regulation (5'UTR global, coding local) and impacting disease traits.
Showing thousands of SNVs, inc. missense, alter translation efficiency, exhibiting two-layer regulation (5'UTR global, coding local) and impacting disease traits.
Taxonomy-aware, disorder-matched benchmarking of phase-separating protein predictors [new]
...reduces shortcut biases in PSP prediction, revealing taxon-specific signals and challenges for IDR-lacking PSPs, guiding model improvement.
...reduces shortcut biases in PSP prediction, revealing taxon-specific signals and challenges for IDR-lacking PSPs, guiding model improvement.
February 12, 2026 at 11:12 PM
Taxonomy-aware, disorder-matched benchmarking of phase-separating protein predictors [new]
...reduces shortcut biases in PSP prediction, revealing taxon-specific signals and challenges for IDR-lacking PSPs, guiding model improvement.
...reduces shortcut biases in PSP prediction, revealing taxon-specific signals and challenges for IDR-lacking PSPs, guiding model improvement.
tensorOmics: Data integration for longitudinal omics data using tensor factorisation [new]
...unifies supervised/unsupervised multi-omic analysis by preserving temporal structure and integrating phenotypic responses.
...unifies supervised/unsupervised multi-omic analysis by preserving temporal structure and integrating phenotypic responses.
February 12, 2026 at 11:11 PM
tensorOmics: Data integration for longitudinal omics data using tensor factorisation [new]
...unifies supervised/unsupervised multi-omic analysis by preserving temporal structure and integrating phenotypic responses.
...unifies supervised/unsupervised multi-omic analysis by preserving temporal structure and integrating phenotypic responses.
Deep learning-based non-invasive profiling of tumor transcriptomes from cell-free DNA for precision oncology [new]
predicts gene expression and pathway enrichment from standard cfDNA sequencing using frag. and nucleosome patterns.
predicts gene expression and pathway enrichment from standard cfDNA sequencing using frag. and nucleosome patterns.
February 12, 2026 at 10:05 PM
Deep learning-based non-invasive profiling of tumor transcriptomes from cell-free DNA for precision oncology [new]
predicts gene expression and pathway enrichment from standard cfDNA sequencing using frag. and nucleosome patterns.
predicts gene expression and pathway enrichment from standard cfDNA sequencing using frag. and nucleosome patterns.
Exploring the mammalian metabolome with DeepMet [new]
unveiling its intricate molecular composition using advanced computational methods.
unveiling its intricate molecular composition using advanced computational methods.
February 12, 2026 at 4:28 PM
Exploring the mammalian metabolome with DeepMet [new]
unveiling its intricate molecular composition using advanced computational methods.
unveiling its intricate molecular composition using advanced computational methods.
A hyperparameter benchmark of VAE-based methods for scRNA-seq batch integration [new]
...investigates architecture settings (latent size, depth/width) across diverse datasets & gene sets, offering practical tuning guidelines.
...investigates architecture settings (latent size, depth/width) across diverse datasets & gene sets, offering practical tuning guidelines.
February 12, 2026 at 4:06 PM
A hyperparameter benchmark of VAE-based methods for scRNA-seq batch integration [new]
...investigates architecture settings (latent size, depth/width) across diverse datasets & gene sets, offering practical tuning guidelines.
...investigates architecture settings (latent size, depth/width) across diverse datasets & gene sets, offering practical tuning guidelines.
Structure-guided analysis and prediction of human E2-E3 ligase pairing specificity [new]
...employs PDB & AlphaFold structures with machine learning to predict E2-E3 pairings, mapping ubiquitination networks.
...employs PDB & AlphaFold structures with machine learning to predict E2-E3 pairings, mapping ubiquitination networks.
February 12, 2026 at 4:04 PM
Structure-guided analysis and prediction of human E2-E3 ligase pairing specificity [new]
...employs PDB & AlphaFold structures with machine learning to predict E2-E3 pairings, mapping ubiquitination networks.
...employs PDB & AlphaFold structures with machine learning to predict E2-E3 pairings, mapping ubiquitination networks.
February 12, 2026 at 11:28 AM
Expanding the Enzymatic Landscape for Polyurethane Degradation of Novel Bacterial Urethanases [new]
struct. characterizes novel bacterial urethanases, detailing carbamate bond hydrolysis mech., advancing PUR biocatalytic recycling.
struct. characterizes novel bacterial urethanases, detailing carbamate bond hydrolysis mech., advancing PUR biocatalytic recycling.
February 12, 2026 at 6:06 AM
Expanding the Enzymatic Landscape for Polyurethane Degradation of Novel Bacterial Urethanases [new]
struct. characterizes novel bacterial urethanases, detailing carbamate bond hydrolysis mech., advancing PUR biocatalytic recycling.
struct. characterizes novel bacterial urethanases, detailing carbamate bond hydrolysis mech., advancing PUR biocatalytic recycling.
AlignPCA-2D: PCA-Reduced Euclidean Vector Alignment for 2D Classification in Cryo-EM [new]
Projects cryo-EM particle images & class representations into a compressed PCA space for assignment via Euclidean distance.
Projects cryo-EM particle images & class representations into a compressed PCA space for assignment via Euclidean distance.
February 12, 2026 at 4:56 AM
AlignPCA-2D: PCA-Reduced Euclidean Vector Alignment for 2D Classification in Cryo-EM [new]
Projects cryo-EM particle images & class representations into a compressed PCA space for assignment via Euclidean distance.
Projects cryo-EM particle images & class representations into a compressed PCA space for assignment via Euclidean distance.
Universal Approximation Theorems for Dynamical Systems with Infinite-Time Horizon Guarantees [new]
...achieved for multistable dynamics (Morse-Smale, cont attractors) ensuring $\varepsilon$-$δ$ closeness for infinite time, and a temporal generalization bound.
...achieved for multistable dynamics (Morse-Smale, cont attractors) ensuring $\varepsilon$-$δ$ closeness for infinite time, and a temporal generalization bound.
February 12, 2026 at 3:52 AM
Universal Approximation Theorems for Dynamical Systems with Infinite-Time Horizon Guarantees [new]
...achieved for multistable dynamics (Morse-Smale, cont attractors) ensuring $\varepsilon$-$δ$ closeness for infinite time, and a temporal generalization bound.
...achieved for multistable dynamics (Morse-Smale, cont attractors) ensuring $\varepsilon$-$δ$ closeness for infinite time, and a temporal generalization bound.
A Dynamical Microscope for Multivariate Oscillatory Signals: Validating Regime Recovery on Shared Manifolds [new]
recovers dynamical laws by converting signals to latent trajectory geometry and flow field metrics, discerning overlapping regimes.
recovers dynamical laws by converting signals to latent trajectory geometry and flow field metrics, discerning overlapping regimes.
February 12, 2026 at 3:50 AM
A Dynamical Microscope for Multivariate Oscillatory Signals: Validating Regime Recovery on Shared Manifolds [new]
recovers dynamical laws by converting signals to latent trajectory geometry and flow field metrics, discerning overlapping regimes.
recovers dynamical laws by converting signals to latent trajectory geometry and flow field metrics, discerning overlapping regimes.