Laura Cantini

Kontext identifies spatial domains and cell-cell communication by integrating ligand-receptor interactions, outperforming existing methods in spatial transcriptomics.

The paper introduces Xpressor, a framework for cross-scale learning in biology, improving cell-type prediction and embedding quality.

This study links transcriptomic data with real-time cell behavior to uncover molecular fate asymmetry and a quiescence cycle in adult neural stem cells.

The paper introduces FLOWSATATE, a trajectory inference method using time-resolved RNA labeling data to model cell fate dynamics and predict future states.

CHAMPOLLION improves multi-omics integration by using regularized optimal transport to align unpaired cells and reveal cross-modal relationships, outperforming existing methods with fewer paired cells.

CIRCE predicts cis-regulatory DNA interactions from single-cell chromatin accessibility data, improving speed and memory use compared to Cicero.

scPRINT-2 is a single-cell foundation model pre-trained on 350 million cells across 16 organisms, achieving state-of-the-art performance in expression denoising, cell embedding, and cell type prediction.

ReCoN models multicellular coordination by integrating cell-cell communication and intracellular regulation to predict cellular responses to perturbations.

This study shows that CENP-A, a centromeric histone variant, epigenetically regulates epithelial-mesenchymal plasticity and heterogeneity in human cells through non-centromeric mechanisms.

MOTL improves multi-omics data factorization for small datasets by transferring knowledge from large datasets, enhancing cancer subtype identification in glioblastoma samples.

scPRINT pre-trains on 50 million cells to predict gene networks, outperforming existing methods in inference and zero-shot tasks.

geneRNIB provides a benchmark for gene regulatory network inference, showing simpler models often outperform complex ones in perturbation-informed metrics.

This study identifies IL-23's role in shaping the functional plasticity of human MAIT cells, revealing its impact on autoimmunity and cytotoxic functions through AP-1 transcription factors.

scConfluence improves single-cell diagonal integration by combining autoencoders with regularized Inverse Optimal Transport, outperforming existing methods in three biological applications.

The paper introduces STORIES, a method for learning cell fate landscapes from spatial transcriptomics data, showing improved spatial coherence in neural regeneration and gliogenesis analyses.

HuMMuS infers regulatory mechanisms from single-cell multi-omics data, improving predictions of TF targets, binding motifs, and regulatory regions.

Mowgli integrates paired single-cell multi-omics data using a combination of NMF and OT, improving clustering and interpretability, and is benchmarked against state-of-the-art methods.

This paper introduces Optimal Transport as a cell-cell similarity metric for single-cell omics data, showing improved clustering performance across multiple datasets.

This study benchmarks four single-cell network inference methods using real data to assess their reproducibility across different conditions and parameters, finding GENIE3 to be the most consistent.

This study benchmarks nine joint multi-omics dimensionality reduction methods using simulated and real cancer data, finding intNMF excels in clustering and MCIA shows consistent performance across contexts.

The paper analyzes Wikipedia's protein functional network using the REGOMAX algorithm to identify evolving hidden communities of biological functions.

The paper uses matrix factorization to analyze molecular mechanisms behind the inverse comorbidity between Alzheimer’s disease and lung cancer, identifying new candidate factors and subtype-specific mechanisms.

The paper introduces a stabilized ICA method that improves reproducibility and interpretability of transcriptomic signals across cancer datasets, and applies it to colorectal cancer.

The paper identifies miRNA clusters driving triple negative breast cancer subtypes using a new computational pipeline and validates their role in promoting the phenotype through proliferation and EMT control.

Moonlight identifies cancer driver genes and their roles by integrating functional analysis and gene regulatory networks from expression data, distinguishing tumor suppressors, oncogenes, and moonlight genes.

The paper defines a set of informative gene signatures from pan-cancer data and introduces InfoSigMap to visualize their compositional and functional redundancies.

This paper introduces a method to determine the optimal number of independent components for reproducible transcriptomic data analysis using ICA stability rankings.