Tommaso Biancalani

SCALLOPS improves processing of optical pooled screens by enabling efficient alignment and analysis of large-scale image data, demonstrating superior performance over existing methods.

The paper introduces PhenoCompass, a model for virtual phenotypic screening that identifies novel PI3K/mTOR inhibitors from a large compound library.

The paper introduces SIGnature, a framework for scoring gene importance using attributions from scRNA-seq foundation models, and demonstrates its utility in identifying novel associations of the MS1 signature with hyperinflammatory conditions.

The paper introduces Decima, a model that predicts cell type- and condition-specific gene expression from DNA sequence using single-cell RNA sequencing data.

This study uses dynamic caging to link live-cell behavior with transcriptional responses across perturbations, revealing new insights into immune and inflammatory gene regulation.

This paper introduces TxConformal, a framework that controls false discoveries in AI-driven therapeutic discovery by adjusting for distribution shifts and providing confidence measures for AI scores.

PaSCient learns patient-level disease representations from scRNA-seq data, enabling disease classification and downstream analyses like dimensionality reduction and patient subgroup discovery.

The paper develops a framework to decode TLS maturation in NSCLC using spatial omics data and deep learning, identifying two distinct spatial ecosystems linked to patient outcomes.

The paper identifies Grem1-expressing fibroblastic reticular cells as a source of FLT3L that supports dendritic cell homeostasis and function in lymph nodes.

Nona is a multimodal masking framework that unifies sequence-to-function models, language models, and generative models for functional genomics, improving predictions and revealing privacy vulnerabilities in ATAC-seq data.

Tangram2 maps cell-to-cell communication by integrating spatial transcriptomics with single-cell RNA sequencing to identify gene expression changes driven by neighboring cell types.

This paper introduces EVFI and DeepEVFI, machine learning methods that infer variant fitness from directed evolution data, identifying tighter binders missed by conventional approaches.

PRESAGE predicts perturbation expression responses using gene embeddings and outperforms existing methods in functional evaluations.

Polygraph assesses synthetic regulatory DNA elements using features like diversity and motif composition to improve gene therapy design.

This paper builds and benchmarks perturbative maps of transcriptional and morphological data, demonstrating their utility for discovering new biology and highlighting the importance of standardized frameworks.

gReLU provides a framework for DNA sequence modeling and design, including tools for training, evaluating, and applying pre-trained models to predict and create regulatory elements.

This study combines high-throughput screening with deep learning to identify novel antibacterial compounds, achieving a 90X higher hit rate than traditional methods.

This paper introduces regLM, a framework using autoregressive language models to design synthetic regulatory DNA elements with desired properties, showing they match biological features of validated CREs.

The paper introduces SCimilarity, a metric learning framework for querying and comparing single-cell RNA-seq data across the human body, revealing commonalities in fibrosis-associated macrophages.

This study evaluates six algorithms for cell type matching in spatial transcriptomics data and presents consensus strategies for improved analysis.

Raman2RNA predicts single-cell RNA expression profiles from live-cell Raman microscopy images using machine learning, achieving high accuracy in tracking cell state transitions during reprogramming.

The paper introduces Tangram, a method that aligns single-nucleus RNA-seq data with spatial data to create an anatomically integrated spatial map of the mouse brain.

The paper creates a comprehensive atlas of mouse primary motor cortex cell types using integrated transcriptomic and epigenomic data from over 600,000 samples.