Alejandro Buendia

BenchRep-T evaluates nine T-cell repertoire methods across four tasks, finding simple baselines competitive with complex models and no single method dominates all tasks.

This study uses Cas12a-based knockout screens to identify genetic interactions in the DNA damage response, revealing synthetic lethal and buffering interactions with therapeutic implications.

Bio-BLIP is a multimodal architecture that integrates DNA, genes, proteins, and text to improve variant interpretation without task-specific fine-tuning, achieving better performance in variant prioritization and target gene prediction.

SpatialProp predicts the effects of multi-gene perturbations on whole tissue sections using spatially resolved single-cell transcriptomics and graph neural networks.

GET is a foundation model that accurately predicts gene expression and regulatory interactions across 213 human cell types using chromatin accessibility and sequence data.

An evaluation of scBERT and scGPT on cell type annotation against L1-regularized logistic regression, with ablations of pre-training and tests of fine-tuning stability.