AI x Bio Discovery
@aixbiobot.bsky.social
Automated discovery of AI x Bio papers, blogs, and news.
Scaling Laws for Masked-Reconstruction Transformers on Single-Cell Transcriptomics [new]
...are systematically studied, revealing power-law scaling in data-rich single-cell transcriptomics, but negligible scaling when data is limited.
...are systematically studied, revealing power-law scaling in data-rich single-cell transcriptomics, but negligible scaling when data is limited.
February 18, 2026 at 3:12 AM
Scaling Laws for Masked-Reconstruction Transformers on Single-Cell Transcriptomics [new]
...are systematically studied, revealing power-law scaling in data-rich single-cell transcriptomics, but negligible scaling when data is limited.
...are systematically studied, revealing power-law scaling in data-rich single-cell transcriptomics, but negligible scaling when data is limited.
Molecular Design beyond Training Data with Novel Extended Objective Functionals of Generative AI Models Driven by Quantum Annealing Computer [new]
integrates quant. annealing with novel NNs to enhance drug-likeness of generated molecules beyond training data.
integrates quant. annealing with novel NNs to enhance drug-likeness of generated molecules beyond training data.
February 18, 2026 at 3:10 AM
Molecular Design beyond Training Data with Novel Extended Objective Functionals of Generative AI Models Driven by Quantum Annealing Computer [new]
integrates quant. annealing with novel NNs to enhance drug-likeness of generated molecules beyond training data.
integrates quant. annealing with novel NNs to enhance drug-likeness of generated molecules beyond training data.
CAMEL: An ECG Language Model for Forecasting Cardiac Events [new]
forecasts future cardiac events by enabling long signal inference via a specialized ECG encoder for cross-understanding with text, trained with LLM methods.
forecasts future cardiac events by enabling long signal inference via a specialized ECG encoder for cross-understanding with text, trained with LLM methods.
February 18, 2026 at 3:08 AM
CAMEL: An ECG Language Model for Forecasting Cardiac Events [new]
forecasts future cardiac events by enabling long signal inference via a specialized ECG encoder for cross-understanding with text, trained with LLM methods.
forecasts future cardiac events by enabling long signal inference via a specialized ECG encoder for cross-understanding with text, trained with LLM methods.
MRC-GAT: A Meta-Relational Copula-Based Graph Attention Network for Interpretable Multimodal Alzheimer's Disease Diagnosis [new]
Aligns & combines multimodal feats (via copula trans. & multi-rel. attn.) to overcome fixed graph lims in episodic meta-learning.
Aligns & combines multimodal feats (via copula trans. & multi-rel. attn.) to overcome fixed graph lims in episodic meta-learning.
February 18, 2026 at 3:07 AM
MRC-GAT: A Meta-Relational Copula-Based Graph Attention Network for Interpretable Multimodal Alzheimer's Disease Diagnosis [new]
Aligns & combines multimodal feats (via copula trans. & multi-rel. attn.) to overcome fixed graph lims in episodic meta-learning.
Aligns & combines multimodal feats (via copula trans. & multi-rel. attn.) to overcome fixed graph lims in episodic meta-learning.
MolDeBERTa: Foundational Model for Physicochemical and Structural-Informed Molecular Representation Learning [new]
Robust mol reps from integrating phys-chem&struct biases via novel pretrain objs & byte-BPE tok on large SMILES.
Robust mol reps from integrating phys-chem&struct biases via novel pretrain objs & byte-BPE tok on large SMILES.
February 18, 2026 at 2:02 AM
MolDeBERTa: Foundational Model for Physicochemical and Structural-Informed Molecular Representation Learning [new]
Robust mol reps from integrating phys-chem&struct biases via novel pretrain objs & byte-BPE tok on large SMILES.
Robust mol reps from integrating phys-chem&struct biases via novel pretrain objs & byte-BPE tok on large SMILES.
Towards inferring atomic scale conformation landscape of biomolecules from cryo-electron tomography data [new]
A deep learning framework predicts atom coordinates from cryo-ET subtoms overcoming comp. limits of physics-based fitting.
A deep learning framework predicts atom coordinates from cryo-ET subtoms overcoming comp. limits of physics-based fitting.
February 18, 2026 at 2:00 AM
Towards inferring atomic scale conformation landscape of biomolecules from cryo-electron tomography data [new]
A deep learning framework predicts atom coordinates from cryo-ET subtoms overcoming comp. limits of physics-based fitting.
A deep learning framework predicts atom coordinates from cryo-ET subtoms overcoming comp. limits of physics-based fitting.
QuantCell: machine learning based cell annotation from qualitative and quantitative imaging profiles [new]
...leverages quantitative data for robust annotation in complex tissues, identifying rare cell types via ML and FDR.
...leverages quantitative data for robust annotation in complex tissues, identifying rare cell types via ML and FDR.
February 18, 2026 at 1:59 AM
QuantCell: machine learning based cell annotation from qualitative and quantitative imaging profiles [new]
...leverages quantitative data for robust annotation in complex tissues, identifying rare cell types via ML and FDR.
...leverages quantitative data for robust annotation in complex tissues, identifying rare cell types via ML and FDR.
Diffusion Probabilistic Models for Missing-Wedge Correction in Cryo-Electron Tomography [new]
are adapted to generate missing 2D tilt images, directly addressing distortions in 3D tomogram reconstructions.
are adapted to generate missing 2D tilt images, directly addressing distortions in 3D tomogram reconstructions.
February 18, 2026 at 1:04 AM
Diffusion Probabilistic Models for Missing-Wedge Correction in Cryo-Electron Tomography [new]
are adapted to generate missing 2D tilt images, directly addressing distortions in 3D tomogram reconstructions.
are adapted to generate missing 2D tilt images, directly addressing distortions in 3D tomogram reconstructions.
TITAN-BBB: Predicting BBB Permeability using Multi-Modal Deep-Learning Models [new]
integrates diverse data (tab, img, txt) via attention mech for computational BBB permeability prediction, trained on an extensive aggregated dataset.
integrates diverse data (tab, img, txt) via attention mech for computational BBB permeability prediction, trained on an extensive aggregated dataset.
February 18, 2026 at 12:55 AM
TITAN-BBB: Predicting BBB Permeability using Multi-Modal Deep-Learning Models [new]
integrates diverse data (tab, img, txt) via attention mech for computational BBB permeability prediction, trained on an extensive aggregated dataset.
integrates diverse data (tab, img, txt) via attention mech for computational BBB permeability prediction, trained on an extensive aggregated dataset.
A Discrete Language of Protein Words for Functional Discovery and Design [new]
...derived from discretizing protein space into an evolutionary vocabulary, capturing higher-order signals for bio discovery and programmable engineering.
...derived from discretizing protein space into an evolutionary vocabulary, capturing higher-order signals for bio discovery and programmable engineering.
February 17, 2026 at 10:07 PM
A Discrete Language of Protein Words for Functional Discovery and Design [new]
...derived from discretizing protein space into an evolutionary vocabulary, capturing higher-order signals for bio discovery and programmable engineering.
...derived from discretizing protein space into an evolutionary vocabulary, capturing higher-order signals for bio discovery and programmable engineering.
MicrowellMicrofluidicsMiner (M3): Leverage Large Language Model Agents for Knowledge Mining of Microwell Microfluidics [new]
Auto-extract microwell array design params from literature, build database for accelerated microfluidic dev.
Auto-extract microwell array design params from literature, build database for accelerated microfluidic dev.
February 17, 2026 at 10:05 PM
MicrowellMicrofluidicsMiner (M3): Leverage Large Language Model Agents for Knowledge Mining of Microwell Microfluidics [new]
Auto-extract microwell array design params from literature, build database for accelerated microfluidic dev.
Auto-extract microwell array design params from literature, build database for accelerated microfluidic dev.
Wavelet-Domain Multi-Representation and Ensemble Learning for Automated ECG Analysis [new]
... leverages scalograms & phasograms from CWT with DL & ensemble-fusion for interpretable, generalizable 5-superclass ECG classification.
... leverages scalograms & phasograms from CWT with DL & ensemble-fusion for interpretable, generalizable 5-superclass ECG classification.
February 17, 2026 at 7:58 PM
Wavelet-Domain Multi-Representation and Ensemble Learning for Automated ECG Analysis [new]
... leverages scalograms & phasograms from CWT with DL & ensemble-fusion for interpretable, generalizable 5-superclass ECG classification.
... leverages scalograms & phasograms from CWT with DL & ensemble-fusion for interpretable, generalizable 5-superclass ECG classification.
February 17, 2026 at 7:15 PM
Evaluating Single-Cell Perturbation Response Models Is Far from Straightforward [new]
Flawed metrics misrepresenting model performance, esp. in high-dimensional data, needs robust guidelines for reliable virt-cell model development.
Flawed metrics misrepresenting model performance, esp. in high-dimensional data, needs robust guidelines for reliable virt-cell model development.
February 17, 2026 at 7:14 PM
Evaluating Single-Cell Perturbation Response Models Is Far from Straightforward [new]
Flawed metrics misrepresenting model performance, esp. in high-dimensional data, needs robust guidelines for reliable virt-cell model development.
Flawed metrics misrepresenting model performance, esp. in high-dimensional data, needs robust guidelines for reliable virt-cell model development.
February 17, 2026 at 6:50 PM
RNAiSpline: A Deep learning model for siRNA efficacy prediction [new]
...overcomes data scarcity and generalization challenges with self-supervised pretraining, KAN, CNN, and Transformer Encoder for robust prediction.
...overcomes data scarcity and generalization challenges with self-supervised pretraining, KAN, CNN, and Transformer Encoder for robust prediction.
February 17, 2026 at 6:49 PM
RNAiSpline: A Deep learning model for siRNA efficacy prediction [new]
...overcomes data scarcity and generalization challenges with self-supervised pretraining, KAN, CNN, and Transformer Encoder for robust prediction.
...overcomes data scarcity and generalization challenges with self-supervised pretraining, KAN, CNN, and Transformer Encoder for robust prediction.
Fast-cWDM Brain MRI: Fast Conditional Wavelet Diffusion Model for Synthesis Brain MRI Modality [new]
combines Fast-DDPM w/ wavelets, reducing steps and using wavelet inputs for rapid missing MRI modality synthesis frm existing scans.
combines Fast-DDPM w/ wavelets, reducing steps and using wavelet inputs for rapid missing MRI modality synthesis frm existing scans.
February 17, 2026 at 6:47 PM
Fast-cWDM Brain MRI: Fast Conditional Wavelet Diffusion Model for Synthesis Brain MRI Modality [new]
combines Fast-DDPM w/ wavelets, reducing steps and using wavelet inputs for rapid missing MRI modality synthesis frm existing scans.
combines Fast-DDPM w/ wavelets, reducing steps and using wavelet inputs for rapid missing MRI modality synthesis frm existing scans.
ML-guided robotic microinjection of single neurons in human brain organoids [new]
...systematically targets and injects single neurons in dense organoid tissue, enabling high-throughput manipulation for brain development studies.
...systematically targets and injects single neurons in dense organoid tissue, enabling high-throughput manipulation for brain development studies.
February 17, 2026 at 8:52 AM
ML-guided robotic microinjection of single neurons in human brain organoids [new]
...systematically targets and injects single neurons in dense organoid tissue, enabling high-throughput manipulation for brain development studies.
...systematically targets and injects single neurons in dense organoid tissue, enabling high-throughput manipulation for brain development studies.
February 17, 2026 at 7:07 AM
Joint Learning of Drug-Drug Combination and Drug-DrugInteraction via Coupled Tensor-Tensor Factorization with SideInformation [new]
models DD & DDI as coupled tensors, using diverse drug sim side info for incompleteness & sparsity.
models DD & DDI as coupled tensors, using diverse drug sim side info for incompleteness & sparsity.
February 17, 2026 at 7:06 AM
Joint Learning of Drug-Drug Combination and Drug-DrugInteraction via Coupled Tensor-Tensor Factorization with SideInformation [new]
models DD & DDI as coupled tensors, using diverse drug sim side info for incompleteness & sparsity.
models DD & DDI as coupled tensors, using diverse drug sim side info for incompleteness & sparsity.
jazzPanda: A hybrid approach to find spatial markergenes in imaging-based spatial transcriptomics data [new]
finds spatial marker genes by pseudobulking gene detections & cells in clusters for LM input, accnting for spatial distr.
finds spatial marker genes by pseudobulking gene detections & cells in clusters for LM input, accnting for spatial distr.
February 17, 2026 at 7:04 AM
jazzPanda: A hybrid approach to find spatial markergenes in imaging-based spatial transcriptomics data [new]
finds spatial marker genes by pseudobulking gene detections & cells in clusters for LM input, accnting for spatial distr.
finds spatial marker genes by pseudobulking gene detections & cells in clusters for LM input, accnting for spatial distr.
Programmable RNA detection generates DNA barcodes for multiplexed phage-host interaction screening [new]
...coupling toehold switches & retron RT to transduce transient RNA to stable DNA records for quant. host suscept. profiling.
...coupling toehold switches & retron RT to transduce transient RNA to stable DNA records for quant. host suscept. profiling.
February 17, 2026 at 5:17 AM
Programmable RNA detection generates DNA barcodes for multiplexed phage-host interaction screening [new]
...coupling toehold switches & retron RT to transduce transient RNA to stable DNA records for quant. host suscept. profiling.
...coupling toehold switches & retron RT to transduce transient RNA to stable DNA records for quant. host suscept. profiling.
Ion Mobility Mass Spectrometry Guided Modeling with AlphaFold and Rosetta Improves Protein Complex Structure Prediction [new]
Using exp coll cross-section data guiding docking of predicted subunits, refining protein complex assembly.
Using exp coll cross-section data guiding docking of predicted subunits, refining protein complex assembly.
February 17, 2026 at 4:55 AM
Ion Mobility Mass Spectrometry Guided Modeling with AlphaFold and Rosetta Improves Protein Complex Structure Prediction [new]
Using exp coll cross-section data guiding docking of predicted subunits, refining protein complex assembly.
Using exp coll cross-section data guiding docking of predicted subunits, refining protein complex assembly.
Learning fragment-based segmentation of binding sites from molecular dynamics: a proof-of-concept on cardiac myosin. [new]
DL segments MD binding sites into frag-spec regions. Monitors apo ligand binding for docking & design.
DL segments MD binding sites into frag-spec regions. Monitors apo ligand binding for docking & design.
February 17, 2026 at 4:11 AM
Learning fragment-based segmentation of binding sites from molecular dynamics: a proof-of-concept on cardiac myosin. [new]
DL segments MD binding sites into frag-spec regions. Monitors apo ligand binding for docking & design.
DL segments MD binding sites into frag-spec regions. Monitors apo ligand binding for docking & design.
Exploring the limits of pre-trained embeddings in machine-guided protein design: a case study on predicting AAV vector viability [new]
Pre-trained protein embeddings require task-specific fine-tuning to capture localized mutations in bioengineering datasets.
Pre-trained protein embeddings require task-specific fine-tuning to capture localized mutations in bioengineering datasets.
February 17, 2026 at 4:10 AM
Exploring the limits of pre-trained embeddings in machine-guided protein design: a case study on predicting AAV vector viability [new]
Pre-trained protein embeddings require task-specific fine-tuning to capture localized mutations in bioengineering datasets.
Pre-trained protein embeddings require task-specific fine-tuning to capture localized mutations in bioengineering datasets.