Motivation: Controlling the outcomes of CRISPR editing is crucial for the success of gene therapy. Since donor template-based editing is often inefficient, alternative strategies have emerged that leverage mutagenic end-joining repair instead. Existing machine learning models can accurately predict end-joining repair outcomes, however: generalisability beyond the specific cell line used for training remains a challenge, and interpretability is typically limited by suboptimal feature representation and model architecture. Results: We propose X-CRISP, a flexible and interpretable neural network for predicting repair outcome frequencies based on a minimal set of outcome and sequence features, including microhomologies (MH). Outperforming prior models on detailed and aggregate outcome predictions, X-CRISP prioritised MH location over MH sequence properties such as GC content for deletion outcomes. Through transfer learning, we adapted X-CRISP pre-trained on wild-type mESC data to target human cell lines K562, HAP1, U2OS, and mESC lines with altered DNA repair function. Adapted X-CRISP models improved over direct training on target data from as few as 50 samples, suggesting that this strategy could be leveraged to build models for new domains using a fraction of the data required to train models from scratch. Availability: An implementation of X-CRISP is available at https://github.com/joanagoncalveslab/xcrisp.@en

Motivation Understanding the factors involved in DNA double-strand break (DSB) repair is crucial for the development of targeted anti-cancer therapies, yet the roles of many genes remain unclear. Recent studies show that perturbations of certain genes can alter the distribution of sequence-specific mutations left behind after DSB repair. This suggests that genome-wide screening could reveal novel DSB repair factors by identifying genes whose perturbation causes the mutational distribution spectra observed at a given DSB site to deviate significantly from the wild-type. However, designing proper controls for a genome-wide perturbation screen could be challenging. We explore the idea that a genome-wide screen might allow us to forgo the use of traditional non-targeting controls by reframing the analysis as an outlier detection problem, assuming that most genes have minimal influence on DSB repair.

Results We propose MUSICiAn (Mutational Signature Catalogue Analysis), a compositional data analysis method that ranks gene perturbation-specific mutational spectra without controls by measuring deviations from the central tendency in the distributions of all spectra. We show that MUSICiAn can effectively estimate pseudo-controls for the existing Repair-seq dataset, screening 476 genes and 60 non-targeting controls. We further apply MUSICiAn to a genome-wide dataset profiling mutational outcomes induced by CRISPR-Cas9 at three target sites across cells with individual perturbations of 18,406 genes. MUSICiAn successfully recovers known genes, highlights the spliceosome as a lesser-appreciated player in DSB repair, and reveals candidates for further investigation.

Availability github.com/joanagoncalveslab/MUSICiAn.@en

Correction to: Nature Cell Biologyhttps://doi.org/10.1038/s41556-023-01194-w. Published online 19 July 2023. In the version of this article originally published, the name of Tianyang Xu was misspelled as Tiangyang Xu. The name has been corrected in the HTML and PDF versions of the article.

@en

There is a need to define regions of gene activation or repression that control human kidney cells in states of health, injury, and repair to understand the molecular pathogenesis of kidney disease and design therapeutic strategies. Comprehensive integration of gene expression with epigenetic features that define regulatory elements remains a significant challenge. We measure dual single nucleus RNA expression and chromatin accessibility, DNA methylation, and H3K27ac, H3K4me1, H3K4me3, and H3K27me3 histone modifications to decipher the chromatin landscape and gene regulation of the kidney in reference and adaptive injury states. We establish a spatially-anchored epigenomic atlas to define the kidney’s active, silent, and regulatory accessible chromatin regions across the genome. Using this atlas, we note distinct control of adaptive injury in different epithelial cell types. A proximal tubule cell transcription factor network of ELF3, KLF6, and KLF10 regulates the transition between health and injury, while in thick ascending limb cells this transition is regulated by NR2F1. Further, combined perturbation of ELF3, KLF6, and KLF10 distinguishes two adaptive proximal tubular cell subtypes, one of which manifested a repair trajectory after knockout. This atlas will serve as a foundation to facilitate targeted cell-specific therapeutics by reprogramming gene regulatory networks.@en

In recent years Large Language Models (LLMs) have been successfully adapted to the field of Genomics, as shown by models such as DNABERT, DNABERT-2, and Nucleotide Transformer. Despite this, their application in the challenging field of Cancer Genomics remains unexplored. This thesis examines whether cancer genome analysis can benefit from large, pre-trained Transformer-based models, specifically focusing on the newly introduced HyenaDNA architecture. In HyenaDNA, traditional Attention Layers are replaced by so-called Hyena Filters, which consist of recursions of an element-wise multiplicative gating and a long convolution, allowing for the processing of longer sequences while maintaining single-base resolution, and achieving a subquadratic computational cost, aligning well with the specific needs of Cancer Genomics. This study begins by assessing HyenaDNA's capabilities to represent genomic data. Using UMAP, we visualized embeddings computed by a HyenaDNA model pre-trained on the Human Reference Genome, showing that the model effectively learns to distinguish between different genomic regions. Following this, we fine-tuned pre-trained HyenaDNA models on the novel task of classifying cancer types directly from the mutated sequences. Given the lack of existing datasets tailored to our specific application, we built a custom dataset combining data from the DepMap dataset and the Human Reference Genome. Comparing performances obtained in various experimental settings, we conclude that this approach shows significant potential to be effectively employed in Cancer Genomics.

The Human BioMolecular Atlas Program (HuBMAP) aims to create a multi-scale spatial atlas of the healthy human body at single-cell resolution by applying advanced technologies and disseminating resources to the community. As the HuBMAP moves past its first phase, creating ontologies, protocols and pipelines, this Perspective introduces the production phase: the generation of reference spatial maps of functional tissue units across many organs from diverse populations and the creation of mapping tools and infrastructure to advance biomedical research.

@en

Motivation

Anti-cancer therapies based on synthetic lethality (SL) exploit tumour vulnerabilities for treatment with reduced side effects, by targeting a gene that is jointly essential with another whose function is lost. Computational prediction is key to expedite SL screening, yet existing methods are vulnerable to prevalent selection bias in SL data and reliant on cancer or tissue type-specific omics, which can be scarce. Notably, sequence similarity remains underexplored as a proxy for related gene function and joint essentiality.
Results

We propose ELISL, Early–Late Integrated SL prediction with forest ensembles, using context-free protein sequence embeddings and context-specific omics from cell lines and tissue. Across eight cancer types, ELISL showed superior robustness to selection bias and recovery of known SL genes, as well as promising cross-cancer predictions. Co-occurring mutations in a BRCA gene and ELISL-predicted pairs from the HH, FGF, WNT, or NEIL gene families were associated with longer patient survival times, revealing therapeutic potential.@en

Motivation
Synthetic lethality (SL) between two genes occurs when simultaneous loss of function leads to cell death. This holds great promise for developing anti-cancer therapeutics that target synthetic lethal pairs of endogenously disrupted genes. Identifying novel SL relationships through exhaustive experimental screens is challenging, due to the vast number of candidate pairs. Computational SL prediction is therefore sought to identify promising SL gene pairs for further experimentation. However, current SL prediction methods lack consideration for generalizability in the presence of selection bias in SL data.
Results
We show that SL data exhibit considerable gene selection bias. Our experiments designed to assess the robustness of SL prediction reveal that models driven by the topology of known SL interactions (e.g. graph, matrix factorization) are especially sensitive to selection bias. We introduce selection bias-resilient synthetic lethality (SBSL) prediction using regularized logistic regression or random forests. Each gene pair is described by 27 molecular features derived from cancer cell line, cancer patient tissue and healthy donor tissue samples. SBSL models are built and tested using approximately 8000 experimentally derived SL pairs across breast, colon, lung and ovarian cancers. Compared to other SL prediction methods, SBSL showed higher predictive performance, better generalizability and robustness to selection bias. Gene dependency, quantifying the essentiality of a gene for cell survival, contributed most to SBSL predictions. Random forests were superior to linear models in the absence of dependency features, highlighting the relevance of mutual exclusivity of somatic mutations, co-expression in healthy tissue and differential expression in tumour samples.
Availability and implementation
https://github.com/joanagoncalveslab/sbsl
Supplementary information
Supplementary data are available at Bioinformatics online.@en

Understanding the impact of guide RNA (gRNA) and genomic locus on CRISPR-Cas9 activity is crucial to design effective gene editing assays. However, it is challenging to profile Cas9 activity in the endogenous cellular environment. Here we leverage our TRIP technology to integrate ~ 1k barcoded reporter genes in the genomes of mouse embryonic stem cells. We target the integrated reporters (IRs) using RNA-guided Cas9 and characterize induced mutations by sequencing. We report that gRNA-sequence and IR locus explain most variation in mutation efficiency. Predominant insertions of a gRNA-specific nucleotide are consistent with template-dependent repair of staggered DNA ends with 1-bp 5′ overhangs. We confirm that such staggered ends are induced by Cas9 in mouse pre-B cells. To explain observed insertions, we propose a model generating primarily blunt and occasionally staggered DNA ends. Mutation patterns indicate that gRNA-sequence controls the fraction of staggered ends, which could be used to optimize Cas9-based insertion efficiency.

@en

Background
Understanding the relationship between diseases based on the underlying biological mechanisms is one of the greatest challenges in modern biology and medicine. Exploring disease-disease associations by using system-level biological data is expected to improve our current knowledge of disease relationships, which may lead to further improvements in disease diagnosis, prognosis and treatment.
Results
We took advantage of diverse biological data including disease-gene associations and a large-scale molecular network to gain novel insights into disease relationships. We analysed and compared four publicly available disease-gene association datasets, then applied three disease similarity measures, namely annotation-based measure, function-based measure and topology-based measure, to estimate the similarity scores between diseases. We systematically evaluated disease associations obtained by these measures against a statistical measure of comorbidity which was derived from a large number of medical patient records. Our results show that the correlation between our similarity measures and comorbidity scores is substantially higher than expected at random, confirming that our similarity measures are able to recover comorbidity associations. We also demonstrated that our predicted disease associations correlated with disease associations generated from genome-wide association studies significantly higher than expected at random. Furthermore, we evaluated our predicted disease associations via mining the literature on PubMed, and presented case studies to demonstrate how these novel disease associations can be used to enhance our current knowledge of disease relationships.
Conclusions
We present three similarity measures for predicting disease associations. The strong correlation between our predictions and known disease associations demonstrates the ability of our measures to provide novel insights into disease relationships.
@en

The YEASTRACT (http://www.yeastract.com) information system is a tool for the analysis and prediction of transcription regulatory associations in Saccharomyces cerevisiae. Last updated in June 2013, this database contains over 200 000 regulatory associations between transcription factors (TFs) and target genes, including 326 DNA binding sites for 113 TFs. All regulatory associations stored in YEASTRACT were revisited and new information was added on the experimental conditions in which those associations take place and on whether the TF is acting on its target genes as activator or repressor. Based on this information, new queries were developed allowing the selection of specific environmental conditions, experimental evidence or positive/negative regulatory effect. This release further offers tools to rank the TFs controlling a gene or genome-wide response by their relative importance, based on (i) the percentage of target genes in the data set; (ii) the enrichment of the TF regulon in the data set when compared with the genome; or (iii) the score computed using the TFRank system, which selects and prioritizes the relevant TFs by walking through the yeast regulatory network. We expect that with the new data and services made available, the system will continue to be instrumental for yeast biologists and systems biology researchers.@en

Identifying patterns in temporal data is key to uncover meaningful relationships in diverse domains, from stock trading to social interactions. Also of great interest are clinical and biological applications, namely monitoring patient response to treatment or characterizing activity at the molecular level. In biology, researchers seek to gain insight into gene functions and dynamics of biological processes, as well as potential perturbations of these leading to disease, through the study of patterns emerging from gene expression time series. Clustering can group genes exhibiting similar expression profiles, but focuses on global patterns denoting rather broad, unspecific responses. Biclustering reveals local patterns, which more naturally capture the intricate collaboration between biological players, particularly under a temporal setting. Despite the general biclustering formulation being NP-hard, considering specific properties of time series has led to efficient solutions for the discovery of temporally aligned patterns. Notably, the identification of biclusters with time-lagged patterns, suggestive of transcriptional cascades, remains a challenge due to the combinatorial explosion of delayed occurrences. Herein, we propose LateBiclustering, a sensible heuristic algorithm enabling a polynomial rather than exponential time solution for the problem. We show that it identifies meaningful time-lagged biclusters relevant to the response of Saccharomyces cerevisiae to heat stress.@en

Identifying patterns in temporal data supports complex analyses in several domains, including stock markets (finance) and social interactions (social science). Clinical and biological applications, such as monitoring patient response to treatment or characterizing activity at the molecular level, are also of interest. In particular, researchers seek to gain insight into the dynamics of biological processes, and potential perturbations of these leading to disease, through the discovery of patterns in time series gene expression data. For many years, clustering has remained the standard technique to group genes exhibiting similar response profiles. However, clustering defines similarity across all time points, focusing on global patterns which tend to characterize rather broad and unspecific responses. It is widely believed that local patterns offer additional insight into the underlying intricate events leading to the overall observed behavior. Efficient biclustering algorithms have been devised for the discovery of temporally aligned local patterns in gene expression time series, but the extraction of time-lagged patterns remains a challenge due to the combinatorial explosion of pattern occurrence combinations when delays are considered. We present heuristic approaches enabling polynomial rather than exponential time solutions for the problem.

@en

Explaining regulatory mechanisms is crucial to understand complex cellular responses leading to system perturbations. Some strategies reverse engineer regulatory interactions from experimental data, while others identify functional regulatory units (modules) under the assumption that biological systems yield a modular organization. Most modular studies focus on network structure and static properties, ignoring that gene regulation is largely driven by stimulus-response behavior. Expression time series are key to gain insight into dynamics, but have been insufficiently explored by current methods, which often (1) apply generic algorithms unsuited for expression analysis over time, due to inability to maintain the chronology of events or incorporate time dependency; (2) ignore local patterns, abundant in most interesting cases of transcriptional activity; (3) neglect physical binding or lack automatic association of regulators, focusing mainly on expression patterns; or (4) limit the discovery to a predefined number of modules. We propose Regulatory Snapshots, an integrative mining approach to identify regulatory modules over time by combining transcriptional control with response, while overcoming the above challenges. Temporal biclustering is first used to reveal transcriptional modules composed of genes showing coherent expression profiles over time. Personalized ranking is then applied to prioritize prominent regulators targeting the modules at each time point using a network of documented regulatory associations and the expression data. Custom graphics are finally depicted to expose the regulatory activity in a module at consecutive time points (snapshots). Regulatory Snapshots successfully unraveled modules underlying yeast response to heat shock and human epithelial-to-mesenchymal transition, based on regulations documented in the YEASTRACT and JASPAR databases, respectively, and available expression data. Regulatory players involved in functionally enriched processes related to these biological events were identified. Ranking scores further suggested ability to discern the primary role of a gene (target or regulator). Prototype is available at: http://kdbio.inesc-id.pt/software/regulatorysnapshots.@en

Disease gene prioritization aims to suggest potential implications of genes in disease susceptibility. Often accomplished in a guilt-by-association scheme, promising candidates are sorted according to their relatedness to known disease genes. Network-based methods have been successfully exploiting this concept by capturing the interaction of genes or proteins into a score. Nonetheless, most current approaches yield at least some of the following limitations: (1) networks comprise only curated physical interactions leading to poor genome coverage and density, and bias toward a particular source; (2) scores focus on adjacencies (direct links) or the most direct paths (shortest paths) within a constrained neighborhood around the disease genes, ignoring potentially informative indirect paths; (3) global clustering is widely applied to partition the network in an unsupervised manner, attributing little importance to prior knowledge; (4) confidence weights and their contribution to edge differentiation and ranking reliability are often disregarded. We hypothesize that network-based prioritization related to local clustering on graphs and considering full topology of weighted gene association networks integrating heterogeneous sources should overcome the above challenges. We term such a strategy Interactogeneous. We conducted cross-validation tests to assess the impact of network sources, alternative path inclusion and confidence weights on the prioritization of putative genes for 29 diseases. Heat diffusion ranking proved the best prioritization method overall, increasing the gap to neighborhood and shortest paths scores mostly on single source networks. Heterogeneous associations consistently delivered superior performance over single source data across the majority of methods. Results on the contribution of confidence weights were inconclusive. Finally, the best Interactogeneous strategy, heat diffusion ranking and associations from the STRING database, was used to prioritize genes for Parkinson’s disease. This method effectively recovered known genes and uncovered interesting candidates which could be linked to pathogenic mechanisms of the disease.@en

Transcription Factors (TFs) control transcription by binding to specific sites in the promoter regions of the target genes, which can be modelled by structured motifs. In this paper we propose AliBiMotif, a method combining sequence alignment and a biclustering approach based on efficient string matching techniques using suffix trees to unravel approximately conserved sets of blocks (structured motifs) while straightforwardly disregarding non-conserved stretches in-between. The ability to ignore the width of non-conserved regions is a major advantage of the proposed method over other motif finders, as the lengths of the binding sites are usually easier to estimate than the separating distances.@en

PINTA (available at http://www.esat.kuleuven.be/pinta/ ; this web site is free and open to all users and there is no login requirement) is a web resource for the prioritization of candidate genes based on the differential expression of their neighborhood in a genome-wide protein–protein interaction network. Our strategy is meant for biological and medical researchers aiming at identifying novel disease genes using disease specific expression data. PINTA supports both candidate gene prioritization (starting from a user defined set of candidate genes) as well as genome-wide gene prioritization and is available for five species (human, mouse, rat, worm and yeast). As input data, PINTA only requires disease specific expression data, whereas various platforms (e.g. Affymetrix) are supported. As a result, PINTA computes a gene ranking and presents the results as a table that can easily be browsed and downloaded by the user. @en