Stefano Lonardi

This paper introduces LangAAI, a sequence-only model that improves antibody CDR design by conditioning on antigen sequences, showing better performance than existing methods without requiring 3D structures.

The paper compares age acceleration across eight cancer types using epigenetic clocks and finds higher age acceleration in tumor tissues, with notable exceptions in the uterus cohort.

The paper assembles the genome of Pavon 76, a Green Revolution wheat cultivar, providing a chromosome-scale reference for CIMMYT-derived wheat and highlighting structural divergence from the standard reference genome.

SIRIUS generates maximally divergent DNA sequences for a given protein, reducing shared subsequences compared to existing methods.

The paper introduces a protein language model fine-tuned to predict antigen-specific B cell receptors from immune repertoires, showing effectiveness across multiple antigens.

ALLEGRO designs universal CRISPR sgRNA libraries targeting multiple fungal species, validated experimentally across three fungi and extended to others.

This study uses a large language model to predict differentially methylated cytosines in TET and DNMT3 knockout mutants in mice and human embryonic stem cells, identifying significant motifs associated with these predictions.

This paper introduces Ab-Affinity, a large language model that generates high-affinity antibodies against SARS-CoV-2 spike protein peptides using combinatorial optimization and AI.

This study benchmarks eleven HiFi assemblers on eukaryotic genomes and metagenomes, finding Hifiasm superior for complex genomes and Hifiasm-meta for metagenomes.

acCRISPR improves CRISPR screen accuracy by correcting sgRNA activity, identifying essential genes in yeast under various conditions.

RAmbler improves de novo genome assembly of complex repetitive regions using high-quality long reads, outperforming existing tools on synthetic and real data.

The paper presents de novo genome assemblies and a pan-genome analysis of six cowpea accessions, identifying core and noncore genes and their functional enrichment.

This paper predicts DNA methylation for CG, CHG, and CHH contexts in five plant species using sequence and methylation data, showing improved accuracy with gene annotations.

The paper develops DeepGuide, a deep learning model to predict high activity CRISPR-Cas9 and -Cas12a guides in Yarrowia lipolytica using genome-wide functional screens.

MO_SCPLOWUSTACHEC_SCPLOW detects chromatin loops from Hi-C and Micro-C data using scale-space theory, identifying more loops than HiCCUPS and showing strong agreement between datasets.

This paper introduces a deep neural network for predicting essential genes in microbes using only protein sequences, outperforming existing methods that rely on structural or topological features.

OMGS improves genome scaffolding by utilizing multiple optical maps, achieving better contiguity and correctness than existing methods.

This study maps seed coat pattern traits in cowpea to three loci using genomic resources and identifies candidate genes involved in flavonoid biosynthesis.

The paper validates genome-wide CRISPR-Cas9 function in Yarrowia lipolytica by quantifying sgRNA efficiency and identifying high-efficiency guides for 94% of genes.

Selfish identifies differential chromatin interactions by comparing Hi-C contact maps using a self-similarity measure, showing improved accuracy over existing methods.

This study presents a high-quality genome assembly of cowpea, identifies genomic features, and discovers a chromosomal inversion linked to Striga resistance.

The paper develops a MAGIC population for cowpea to dissect genetic traits and improve breeding, identifying trait-associated SNPs and recombination patterns.

This study benchmarks eleven metagenomic classifiers using 846 species and evaluates strategies to reduce false positives in taxonomic classification.

This paper develops genome resources for cowpea to support breeding programs and improve food security in climate-stressed regions.

CLARK-S improves taxonomic classification of short metagenomic reads by enhancing sensitivity without sacrificing precision or speed.

This paper shows that ultra-deep sequencing data can degrade decoding and assembly quality, and proposes slicing data to improve results in BAC clone decoding and de novo assembly.