Ochoa, Idoia

Abstract Deconvolution models are a powerful tool for extracting cell-type-specific information from bulk gene expression profiles. Current methods leverage advanced machine learning models and high-resolution sequencing, like single-cell RNA-sequencing, showing promising results across diverse tissues and conditions. However, they still present important limitations: First, many depend on selecting a robust reference, which can strongly affect the deconvolution. Second, pseudobulk data used for training and real bulk RNA-seq samples often exhibit strong distribution shifts, which are currently unaccounted for. Finally, most deconvolution approaches behave as black boxes, which can compromise the reliability of the results. Here, we present Sweetwater, an adaptive and interpretable autoencoder that efficiently deconvolves bulk samples leveraging multiple classes of reference data. Moreover, we propose an improved way of generating training data from a mixture of FACS-sorted FASTQ files, reducing platform-specific biases and outperforming current single-cell-based references. Furthermore, we introduce a gold standard dataset to facilitate fair and accurate evaluation of deconvolution approaches. Finally, we demonstrate that Sweetwater adapts effectively to deconvolved samples during training, uncovering biologically meaningful patterns and enhancing result’s reliability. Sweetwater is available at https://doi.org/10.6084/m9.figshare.29609180, and we anticipate it will expedite the accurate examination of high-throughput clinical data across diverse applications.

Abstract Grouping gene expression into gene set activity scores (GSAS) provides better biological insights than studying individual genes. However, existing gene set projection methods cannot return representative, robust, and interpretable GSAS. We developed NetActivity, a machine learning framework that generates GSAS based on a sparsely-connected autoencoder, where each neuron in the inner layer represents a gene set. We proposed a three-tier training that yielded representative, robust, and interpretable GSAS. NetActivity model was trained with 1518 GO biological processes terms and KEGG pathways and all GTEx samples. NetActivity generates GSAS robust to the initialization parameters and representative of the original transcriptome, and assigned higher importance to more biologically relevant genes. Moreover, NetActivity returns GSAS with a more consistent definition and higher interpretability than GSVA and hipathia, state-of-the-art gene set projection methods. Finally, NetActivity enables combining bulk RNA-seq and microarray datasets in a meta-analysis of prostate cancer progression, highlighting gene sets related to cell division, key for disease progression. When applied to metastatic prostate cancer, gene sets associated with cancer progression were also altered due to drug resistance, while a classical enrichment analysis identified gene sets irrelevant to the phenotype. NetActivity is publicly available in Bioconductor and GitHub.