Idoia Ochoa

SNMF is a reference-free deconvolution method for spatial transcriptomics that improves accuracy and speed by incorporating spatial correlations and GPU acceleration.

The paper constructs a CAR-T cell functional atlas from over one million cells across 13 studies to link product composition with clinical response and identify sex- and age-dependent effects.

This study develops patient-derived liver biopsy organoids to model alcohol-associated liver disease and identifies disease-specific features and therapeutic responses.

JIND-Multi improves cell annotation in single-cell RNA and ATAC data by transferring labels across multiple datasets, reducing unclassified cells and maintaining accuracy.

This paper introduces a method to jointly reconstruct the evolutionary history of SNVs and CNAs from low-coverage single-cell DNA sequencing data, demonstrating improved accuracy over single-modality approaches.

DeepRBP identifies RNA-binding protein-gene regulation pairs using a deep learning framework and explainability module, validated on known interactions.

NetActivity improves gene set activity scores by using interpretable autoencoders, showing better consistency and biological relevance in prostate cancer analysis.

gMCSpy is a Python package for calculating genetic minimal cut sets in genome-scale metabolic models, offering improved accuracy and speed over existing tools.

GeNNius uses a graph neural network to predict drug-target interactions more efficiently and accurately than existing methods across multiple datasets.

This paper introduces a method to infer clonal trees from ultra-low coverage single-cell DNA sequencing data of tumors, showing improved accuracy in uncovering clonal structure and genotypes.

JIND improves automated cell-type annotation in single-cell RNA-seq by learning a latent space for reliable classification and batch alignment without retraining.

The paper shows that quantizing nanopore quality scores has minimal impact on downstream analysis and suggests using lossy compression for storage savings.

RENANO improves nanopore FASTQ compression by 39.8% for base call sequences using reference-based methods, with overall improvements of 12.7% in some scenarios.

GPress compresses and queries GFF and feature expression files efficiently, achieving significant size reduction without compromising retrieval speed.

The paper introduces SimiC, a method for inferring gene regulatory networks per cell phenotype and comparing them across conditions, revealing regulatory dynamics missed by single GRN approaches.

VEF is a variant filtering tool using ensemble methods that outperforms existing tools like VQSR and HF, with faster runtime and better generalization.

The paper introduces SAMDUDE, a denoising method for aligned genomic data that improves variant calling by reducing false positives and increasing true variant detection in both chromosome and whole genome sequencing data.

MassComp is a lossless compressor for mass spectrometry data that reduces file size by up to 89% compared to uncompressed data and outperforms existing compressors.

The paper introduces E2M, a deep learning framework that predicts methylation patterns from gene expression data in cancer cells, finding that this approach improves parameter estimation and modulates expression assessment.

This paper introduces MPEG-G, an ISO standard for genomic data compression, highlighting its format, decoding process, and interoperability features.

FaStore compresses FASTQ files with lossy modes, achieving better compression ratios than existing methods while maintaining variant calling accuracy.

This paper evaluates the impact of lossy quality score compression on variant calling accuracy using genomic datasets and finds that it can maintain or even improve performance while reducing storage costs.