transformer / foundation model

This paper introduces a learned single-cell spatiotemporal manifold to analyze tissue morphology and dynamics over time.

SNAIL is a named entity recognition framework that identifies bioinformatics software and databases in scientific texts, outperforming existing methods and enabling meta-analysis of tool usage.

This study identifies KDM5D on the Y chromosome as a suppressor of CD8+ T cell antitumor immunity in males, linking it to TCR and cholesterol-exhaustion programs.

The paper structures and engineers deep-sea halophilic PET hydrolases to improve their salt tolerance and catalytic efficiency for industrial applications.

The paper develops DNABERT-Enhancer, a genomic language model for enhancer prediction, achieving high accuracy and identifying allele-specific activity and candidate SNPs.

Bromo, a pairwise transformer, ranks a protein's peptide precursors by expected mass spectrometry response, beating XGBoost, PREGO and PeptideRanger on held-out yeast data, and can be fine-tuned per instrument.

State, a set-based transformer trained on over 100 million perturbed cells, predicts perturbation responses in new cellular contexts better than mean, linear and deep learning baselines.

MetAgeFormer is a transformer-based model pre-trained on metabolomic data to estimate biological age and identify metabolic subtypes associated with disease risk.

Web resource offering predicted monomeric and oligomeric structures for 1,603 M. leprae proteins, with pockets, epitope scores, essentiality and a target priority ranking, plus a nanopore resistance-calling module.

Hi-Cformer models multi-scale chromatin contact maps from single-cell Hi-C data, improving cell type separation and imputation of 3D genome features.

scProtoTransformer provides a scalable reference mapping framework across molecular, cellular, and donor levels using a prototype-based Transformer model with improved interpretability and performance.

OmicsFM is a proteomics foundation model that matches transcriptomic models in performance, revealing complementary biology between proteomics and transcriptomics.

InsilicoCell predicts cellular functional profiles and enables multi-objective virtual drug screening by integrating diverse molecular data and tasks.

WaterFlow predicts ordered water molecule positions in protein structures with sub-angstrom accuracy, outperforming existing methods and improving drug design and crystallographic refinement.

TaHL-PTM improves PTM prediction by addressing tokenization issues in decoder-only PLMs, achieving better performance on residue-level tasks.

This paper introduces CVIT-ESP and ESPFormer, pre-trained Vision Transformer models adapted for EEG-based seizure prediction, achieving higher AUC-ROC than existing models.

CodonMamba is a foundation model for mRNA coding sequence design that achieves state-of-the-art performance and enables programmable codon optimization through inference-time steering.

This paper introduces DualMyo, a Transformer-based model for EMG-to-digit classification, achieving 91% accuracy with lightweight fine-tuning for cross-session adaptation.

The paper introduces a multi-modal framework for biomarker discovery in TNBC using public data, achieving high predictive performance and generating testable hypotheses.

The paper predicts plant morphological complexity using transcription factor data with TabPFN, finding TF family expansion drives complexity evolution.

The paper introduces GB.GeneUnet, a transformer-based model pretrained on multi-species genomes to predict gene expression at megabase-scale contexts, achieving state-of-the-art performance.

The paper introduces a dataset of VHH-Fc antibodies and shows that IgG-trained models outperform VHH-Fc-specific models in predicting developability properties.

This paper designs BMP-2 binders using AI pipelines to control bone regeneration, finding a binder with 2.07 nM affinity for BMP-2.

RegFM improves gene regulation modeling by integrating cis-regulatory elements and trans-acting regulators in human cells, outperforming existing methods in multiple tasks.

This paper develops a compact CNN-Transformer model for decoding handwritten digits from sEMG signals, showing improved performance with few-shot adaptation.

ETAP-CLF uses ESM3 embeddings and transformer attention to classify proteins into binary categories, achieving high AUROC scores across different biological processes.

This study introduces DynGeo-Pheno, a geometric deep learning framework that predicts disease-associated phosphosites and missense mutations by integrating protein dynamics and structural information.

CALFP-MHC improves peptide-MHC binding prediction by using chemically grounded fingerprints and contrastive learning, achieving high AUC and PPV across diverse alleles and class imbalance.

Browser-based ZEISS platform trains U-Net/EfficientNet semantic and Mask2Former instance segmentation models from partial, SAM-assisted labels with fixed hyperparameters; described via two organoid demos, without accuracy benchmarks.

Parnet is a CLIP-seq-based foundation model for RNA sequence representation learning that outperforms existing methods in binding profile prediction and generalizes across tasks.