Mohammad Lotfollahi

CellBridge improves cell classification in pathology images by transferring molecular information, achieving higher accuracy on fine-grained labels.

G2T reconstructs tissue geometry from gene expression data using a generative deep learning model, outperforming existing methods on spatial reconstruction benchmarks.

SQUINT learns discrete cell and niche codes from spatial transcriptomics data using residual vector quantization, outperforming baselines in identification and cross-section integration.

TERRA is a foundation model for spatial transcriptomics that predicts tissue behavior under perturbation and identifies multi-scale biological insights across tissues.

Human Developmental Cell Atlas integrates ~4.6 million cells/nuclei from 4-22 post-conception weeks into 454 refined cell types and maps 114 spatial niches across 6PCW whole-embryo sections.

Tripso, a self-supervised transformer model, identifies gene program-specific embeddings and discovers novel gene programs, revealing biological insights in hematopoietic development and disease.

This study creates a spatially-resolved molecular atlas of human skin and uses deep learning to identify 26 niches, revealing novel immune memory structures in inflammatory skin diseases.

PerturbGen predicts how genetic perturbations affect cellular trajectories across time in human datasets, showing effects on immune responses, hematopoiesis, and skin development.

The paper uses division-resolved single-cell transcriptomics to study T cell activation and identifies gene programmes linked to immune disease by analyzing variant effects on cell division states.

Pertpy provides a modular framework for analyzing large-scale perturbation single-cell experiments, offering tools for metadata annotation and perturbation distance calculations.

This study creates a spatial genomics atlas of human skin fibroblasts in health and disease, identifying six major populations and three disease-specific subtypes.

MintFlow is a generative AI algorithm that predicts how tissue perturbations reprogram cell states in disease, revealing novel cell populations and therapeutic targets in three human conditions.

CellDISECT improves cell type discovery and counterfactual predictions by disentangling covariates and using synthetic data for training.

This paper introduces IMPA, a generative model that predicts cell morphological responses to chemical and genetic perturbations using untreated cell images as input, showing improved performance on structurally related compounds.

This paper introduces a VAE-based model for integrating multimodal single-cell data, showing it can map queries across modalities and improve data imputation and biological discovery.

ArchMap provides a no-code tool for reference-based analysis of single-cell datasets, enabling automated mapping, annotation, and collaboration across reference atlases.

The paper constructs a high-resolution single-cell reference atlas of the human endometrium, identifying cell types, states, and signaling networks, and links them to endometriosis.

The paper introduces MultiMIL, a deep learning framework for phenotype prediction and cell state identification using single-cell data from COVID-19 patients and other datasets.

The paper introduces mvTCR, a multimodal generative model for integrating TCR and transcriptome data to improve analysis of single-cell immune profiling data.

The paper introduces NicheCompass, a graph deep learning method for characterizing cell niches in spatial omics data, showing improved performance in identifying and analyzing niches across multiple biological contexts.

This paper introduces scPoli, a deep learning model for integrating single-cell datasets, enabling multi-scale analysis of population-level data by linking sample and cell-level features.

This paper introduces veloVI, a deep generative model for RNA velocity analysis, showing improved data fit, consistency, and uncertainty quantification compared to existing methods.

The paper introduces treeArches, a framework for building and extending cell-type hierarchies by mapping new datasets to reference atlases and harmonizing annotations across studies.

The paper introduces expiMap, a deep learning model that enables interpretable reference mapping of single-cell transcriptome data by incorporating biologically meaningful gene programs.

MultiCPA predicts multimodal perturbation responses at the single-cell level using a generative model trained on CRISPR perturbation data.

scvi-tools provides a Python package for probabilistic analysis of single-cell omics data, offering standardized interfaces and extensible tools for various analysis tasks.

Squidpy extends Scanpy/AnnData with spatial neighbor graphs and a lazy-loading image container, enabling spatial statistics, ligand-receptor tests, image features, and ~10-fold faster neighborhood enrichment than Giotto.

The paper uses β-VAE to learn disentangled representations for scRNA-seq data and shows improved out-of-distribution prediction compared to existing methods.

The compositional perturbation autoencoder separates drug, dose and cell-type effects in single-cell transcriptomes, predicts unseen combinations, and generates 5,329 unmeasured genetic combinations in silico.

The paper introduces an interpretable autoencoder framework for single-cell RNA-seq data using annotated feature sets, showing improved scalability over factor models.