MOTIVATION: The current diagnostic rate of rare diseases through whole-genome sequencing has stabilized at around 30% on average, highlighting the need for improved computational scores to identify pathogenic variants. In 2019, we developed NCBoost, a supervised-learning approach that mined a comprehensive set of sequence constraint features and proved particularly well suited to identifying high-effect pathogenic non-coding variants in genetic diseases. Since its first release, the substantial increase in the number of variants available for training, as well as the enhanced capacity to detect purifying selection signals from large-scale genome sequencing projects, motivated an update of NCBoost. RESULTS: We implemented NCBoost v2, a pathogenicity score for non-coding single-nucleotide variants, trained on the largest set of curated pathogenic variants in monogenic Mendelian diseases available to date. It leverages conservation features computed from recent large-scale genomic consortia such as Zoonomia and gnomAD, and incorporates recent splice-altering predictive scores. NCBoost v2 outperformed alternative state-of-the-art methods in a variety of scenarii, providing more consistent scores across non-coding genomic regions and fine-tuning the scoring of pathogenic splice-altering variants in Mendelian disease genes. AVAILABILITY AND IMPLEMENTATION: NCBoost v2 software is implemented in Python 3.10 and is freely available under the GNU General Public License Version 3 at https://doi.org/10.5281/zenodo.16029049 and https://github.com/RausellLab/NCBoost-2, together with precomputed scores for the human genome assembly GRCh38.

MOTIVATION: Rare diseases collectively affect 5% of the population. However, fewer than 50% of rare disease patients receive a molecular diagnosis after whole genome sequencing. Supervised machine learning is a valuable approach for the pathogenicity scoring of human genetic variants. However, existing methods are often trained on curated but limited central repositories, resulting in poor accuracy when tested on external cohorts. Yet, large collections of variants generated at hospitals and research institutions remain inaccessible to machine-learning purposes because of privacy and legal constraints. Federated learning (FL) algorithms have been recently developed enabling institutions to collaboratively train models without sharing their local datasets. RESULTS: Here, we present a proof-of-concept study evaluating the effectiveness of FL for the clinical classification of genetic variants. A comprehensive array of diverse FL strategies was assessed for coding and non-coding Single Nucleotide Variants as well as Copy Number Variants. Our results showed that federated models generally achieved comparable or superior performance to traditional centralized learning. In addition, federated models reached a robust generalization to independent sets with smaller data fractions as compared to their centralized model counterparts. Our findings support the adoption of FL to establish secure multi-institutional collaborations in human variant interpretation. AVAILABILITY AND IMPLEMENTATION: All source code required to reproduce the results presented in this article, implemented in Python, is available under the GNU General Public License v3 at https://github.com/RausellLab/FedLearnVar.

UNLABELLED: Cell differentiation processes are achieved through a continuum of hierarchical intermediate cell states that might be captured by single-cell RNA seq. Existing computational approaches for the assessment of cell-state hierarchies from single-cell data can be formalized under a general framework composed of (i) a metric to assess cell-to-cell similarities (with or without a dimensionality reduction step) and (ii) a graph-building algorithm (optionally making use of a cell clustering step). The Sincell R package implements a methodological toolbox allowing flexible workflows under such a framework. Furthermore, Sincell contributes new algorithms to provide cell-state hierarchies with statistical support while accounting for stochastic factors in single-cell RNA seq. Graphical representations and functional association tests are provided to interpret hierarchies. The functionalities of Sincell are illustrated in a real case study, which demonstrates its ability to discriminate noisy from stable cell-state hierarchies. AVAILABILITY AND IMPLEMENTATION: Sincell is an open-source R/Bioconductor package available at http://bioconductor.org/packages/sincell. A detailed manual and a vignette are provided with the package. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Sequencing projects have identified large numbers of rare stop-gain and frameshift variants in the human genome. As most of these are observed in the heterozygous state, they test a gene's tolerance to haploinsufficiency and dominant loss of function. We analyzed the distribution of truncating variants across 16,260 autosomal protein coding genes in 11,546 individuals. We observed 39,893 truncating variants affecting 12,062 genes, which significantly differed from an expectation of 12,916 genes under a model of neutral de novo mutation (p<10-4). Extrapolating this to increasing numbers of sequenced individuals, we estimate that 10.8% of human genes do not tolerate heterozygous truncating variants. An additional 10 to 15% of truncated genes may be rescued by incomplete penetrance or compensatory mutations, or because the truncating variants are of limited functional impact. The study of protein truncating variants delineates the essential genome and, more generally, identifies rare heterozygous variants as an unexplored source of diversity of phenotypic traits and diseases.

Loss-of-function variants in innate immunity genes are associated with Mendelian disorders in the form of primary immunodeficiencies. Recent resequencing projects report that stop-gains and frameshifts are collectively prevalent in humans and could be responsible for some of the inter-individual variability in innate immune response. Current computational approaches evaluating loss-of-function in genes carrying these variants rely on gene-level characteristics such as evolutionary conservation and functional redundancy across the genome. However, innate immunity genes represent a particular case because they are more likely to be under positive selection and duplicated. To create a ranking of severity that would be applicable to innate immunity genes we evaluated 17,764 stop-gain and 13,915 frameshift variants from the NHLBI Exome Sequencing Project and 1,000 Genomes Project. Sequence-based features such as loss of functional domains, isoform-specific truncation and nonsense-mediated decay were found to correlate with variant allele frequency and validated with gene expression data. We integrated these features in a Bayesian classification scheme and benchmarked its use in predicting pathogenic variants against Online Mendelian Inheritance in Man (OMIM) disease stop-gains and frameshifts. The classification scheme was applied in the assessment of 335 stop-gains and 236 frameshifts affecting 227 interferon-stimulated genes. The sequence-based score ranks variants in innate immunity genes according to their potential to cause disease, and complements existing gene-based pathogenicity scores. Specifically, the sequence-based score improves measurement of functional gene impairment, discriminates across different variants in a given gene and appears particularly useful for analysis of less conserved genes.

UNLABELLED: We have implemented in a single package all the features required for extracting, visualizing and manipulating fully conserved positions as well as those with a family-dependent conservation pattern in multiple sequence alignments. The program allows, among other things, to run different methods for extracting these positions, combine the results and visualize them in protein 3D structures and sequence spaces. AVAILABILITY AND IMPLEMENTATION: JDet is a multiplatform application written in Java. It is freely available, including the source code, at http://csbg.cnb.csic.es/JDet. The package includes two of our recently developed programs for detecting functional positions in protein alignments (Xdet and S3Det), and support for other methods can be added as plug-ins. A help file and a guided tutorial for JDet are also available.

MOTIVATION: Current projects for the massive characterization of proteomes are generating protein sequences and structures with unknown function. The difficulty of experimentally determining functionally important sites calls for the development of computational methods. The first techniques, based on the search for fully conserved positions in multiple sequence alignments (MSAs), were followed by methods for locating family-dependent conserved positions. These rely on the functional classification implicit in the alignment for locating these positions related with functional specificity. The next obvious step, still scarcely explored, is to detect these positions using a functional classification different from the one implicit in the sequence relationships between the proteins. Here, we present two new methods for locating functional positions which can incorporate an arbitrary external functional classification which may or may not coincide with the one implicit in the MSA. The Xdet method is able to use a functional classification with an associated hierarchy or similarity between functions to locate positions related to that classification. The MCdet method uses multivariate statistical analysis to locate positions responsible for each one of the functions within a multifunctional family. RESULTS: We applied the methods to different cases, illustrating scenarios where there is a disagreement between the functional and the phylogenetic relationships, and demonstrated their usefulness for the phylogeny-independent prediction of functional positions.