Annalisa Marsico

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.

FENNEC predicts siRNA activity across chemically diverse designs using a neural network ensemble, outperforming existing methods and supporting design optimization.

RIBEX improves RBP prediction by integrating sequence and interactome data, outperforming existing methods on IDR-rich and non-RBD proteins.

This study uses CRISPR-Cas9 to identify functional exons in lncRNAs, finding most exons are non-functional and functionality is enriched at the 5' end.

The paper identifies 121 lncRNA driver mutations across 19 cancer types using 12,631 genomes, showing mutations disrupt RNA-binding protein interactions to promote oncogenesis.

This paper introduces a graph-based topic modelling method for single-cell multi-omic data, showing improved performance in clustering and biological interpretation.

miRXplain predicts miRNA-mRNA interactions using CLIP-L data and hybrid attention transformers, outperforming existing models and revealing isomiR-specific targeting mechanisms.

The paper constructs co-expression networks using gene and isoform data from individuals with and without stress-related psychiatric disorders, revealing distinct network structures and regulatory rewiring.

BioPathNet improves link prediction in biomedical knowledge graphs by using path-based reasoning, outperforming existing methods and identifying novel drug indications and gene interactions.

The paper shows that microRNA activity is essential for early mouse embryos to transition into diapause, with miRNAs regulating gene expression and linking mTOR activity to miRNA biogenesis.

This study benchmarks 11 machine learning methods for predicting in vivo RBP-RNA interactions across hundreds of CLIP-seq datasets and RBPs.

RBPNet predicts CLIP crosslink count distributions from RNA sequences, outperforming existing methods and enabling variant impact scoring.

This paper introduces ChromTransfer, a transfer learning method that improves cell-type specific chromatin accessibility prediction from DNA sequences using pre-trained models.

The paper maps human RBP-viral RNA interactions for SARS-CoV-2 and related coronaviruses using deep learning, identifies variant impacts, and links RBPs to potential biomarkers.

The paper identifies distal and proximal cis-regulatory elements that respond to X-chromosomal dosage and developmental state to regulate Xist expression during XCI in mouse embryonic stem cells.

The paper models lncRNA chromatin (dis-)association using nascent RNA sequencing and identifies factors influencing dissociation and its functional impact.

The paper uses allele-specific single-cell RNA-sequencing to study Xist upregulation and gene silencing during X-chromosome inactivation in mouse embryonic stem cells, identifying Nanog downregulation as a trigger and genetic variation's role in XCI.

The paper predicts Xist-induced gene silencing kinetics using a Random Forest model based on epigenetic and genomic features, identifying key factors like distance to Xist and promoter pre-marking.

The paper develops a graph analysis method combining co-expression and ChIA-PET data to identify lncRNA-mediated regulatory modules in K562 cells, finding enriched disease associations and functional roles for lncRNAs.

TriPepSVM predicts RNA-binding proteins using tripeptide frequencies, outperforming existing methods in human and bacterial species.

pysster is a Python package that trains convolutional neural networks to classify biological sequences by learning sequence and structure motifs, achieving higher accuracy than existing methods.

PureCLIP improves crosslink site detection in CLIP-seq data by modeling truncation patterns and non-specific biases, outperforming existing methods in accuracy and replicate consistency.

The paper identifies A-ROD as a long ncRNA that requires chromatin-release to activate DKK1 transcription by recruiting EBP1 to the promoter.

ssHMM identifies sequence-structure motifs for RNA-binding proteins using a hidden Markov model and Gibbs sampling, outperforming existing tools in accuracy and speed.