Bioinformatics (Oxford, England)[JOURNAL]
Sun
200 papers
RAREsim2: flexible simulation of rare variant genetic data using real haplotypes.
Murphy JI, Barnard R, Null M … +1 more , Hendricks AE
Bioinformatics · 2026 May · PMID 42085453 · Full text
MOTIVATION: Realistic simulated data is critical for advancing methodological development and optimizing study design in genetics research. However, many genetic simulation tools are unable to replicate the distribution...
MOTIVATION: Realistic simulated data is critical for advancing methodological development and optimizing study design in genetics research. However, many genetic simulation tools are unable to replicate the distribution of rare variants or incorporate key genetic information, such as functional annotations and linkage disequilibrium. RAREsim, an accurate rare variant simulation algorithm that uses real genetic haplotypes, was developed to address these limitations. Here, we introduce RAREsim2, an update that provides both streamlined software and new functionalities for simulating individual-level differences (e.g., case-control status, technological or batch effects) and variant-level differences to represent a variety of causal models. RESULTS: We demonstrate RAREsim2's utility with three rare variant association methods (Burden, SKAT, and SKAT-O) across several simulation scenarios, including various genetic ancestries, gene sizes, strengths of association, and proportions of risk variants. Type I Error was maintained and the test with the highest power matched previously known patterns. Importantly, real genetic regions can be simulated to include known variant functions and disease associations. Ultimately, RAREsim2 offers additional flexibility and ease in simulating a multitude of realistic genetic scenarios. AVAILABILITY AND IMPLEMENTATION: The RAREsim2 Python package is publicly available on Github (https://github.com/Hendricks-Research-Team/RAREsim2), PyPI (https://pypi.org/project/raresim/), and Zenodo (https://doi.org/10.5281/zenodo.19442523). Code for the example demonstration can be found at https://github.com/JessMurphy/RAREsim2-demo.
AXOLOTL: an accurate method for detecting aberrant gene expression in rare diseases using coexpression constraints.
Xu W, Shen Y, Liu X … +2 more , Leng F, Liu Y
Bioinformatics · 2026 May · PMID 42083807 · Full text
MOTIVATION: The assessment of aberrant transcription events in rare disease patients holds great promise for enhancing the prioritization of causative genes-a strategy already widely adopted in clinical settings to impro...
MOTIVATION: The assessment of aberrant transcription events in rare disease patients holds great promise for enhancing the prioritization of causative genes-a strategy already widely adopted in clinical settings to improve diagnostic accuracy. Nevertheless, the accurate identification of causal genes remains a substantial challenge. RESULTS: We propose AXOLOTL, a novel ensemble method for identifying aberrant gene expression events in RNA expression matrices. AXOLOTL effectively accounts for gene correlation by incorporating coexpression constraints. We demonstrated the superior performance of AXOLOTL on representative RNA-seq datasets, including those from the GTEx healthy cohort, mitochondrial disease cohorts, and collagen VI-related dystrophy cohorts. Furthermore, we applied AXOLOTL to real-world cases of neurological disorders and demonstrated its ability to accurately identify aberrant gene expression and facilitate the prioritization of pathogenic variants. AVAILABILITY AND IMPLEMENTATION: AXOLOTL is freely available on GitHub (https://github.com/xuwenjian85/axolotl) and Zenodo (https://doi.org/10.5281/zenodo.17940844).
Fault-tolerant pedigree reconstruction from pairwise kinship relations.
Huang EC, Li KA, Narasimhan VM
Bioinformatics · 2026 Jun · PMID 42083796 · Full text
MOTIVATION: Pedigrees reconstructed from biologically related ancient genomes have revealed many insights into (pre)history. To our knowledge, all reported ancient pedigrees have been primarily manually reconstructed, as...
MOTIVATION: Pedigrees reconstructed from biologically related ancient genomes have revealed many insights into (pre)history. To our knowledge, all reported ancient pedigrees have been primarily manually reconstructed, as existing pedigree reconstruction methods are ill-suited for the quality and nature of ancient DNA data. RESULTS: We introduce repare, an open-source software method to automatically reconstruct pedigrees from inferred pairwise kinship relations, which are readily obtainable from ancient genomes. This method reconstructs pedigrees by iteratively incorporating pairwise kinship relations into a set of candidate pedigrees, with pruning and sampling to reduce its search space. It optionally considers supporting information such as haplogroups and skeletal age-at-death estimates. We evaluate this method on a variety of simulated pedigrees with varying error rates and missingness. We also use this method to reconstruct several published pedigrees that were originally manually reconstructed; for one, we present a potential alternative topology. repare optionally incorporates user-inferred pedigree constraints, enabling "human-in-the-loop" reconstruction workflows. Especially when used with these user-inferred constraints, we find that repare represents a powerful and flexible tool for ancient pedigree reconstruction. AVAILABILITY AND IMPLEMENTATION: repare is freely available at https://github.com/Narasimhan-Lab/repare. In addition, source code, benchmark scripts, and benchmark results used in this work are archived at https://doi.org/10.5281/zenodo.19716772.
A framework to infer de novo exonic variants when parental genotypes are missing enhances association studies of autism.
Moon H, Sloofman L, Avila MN … +4 more , Klei L, Devlin B, Buxbaum JD, Roeder K
Bioinformatics · 2026 May · PMID 42082430 · Full text
MOTIVATION: Gene-damaging mutations are highly informative for studies seeking to discover genes underlying developmental disorders. Traditionally, these de novo variants are recognized by evaluating high-quality DNA seq...
MOTIVATION: Gene-damaging mutations are highly informative for studies seeking to discover genes underlying developmental disorders. Traditionally, these de novo variants are recognized by evaluating high-quality DNA sequence from affected offspring and parents. However, when parental sequence is unavailable, methods are required to infer de novo status and use this inference for association studies. RESULTS: We use data from autism spectrum disorder to illustrate and evaluate methods. Separating de novo from rare inherited variants is challenging because the latter are far more common. Using a classifier for unbalanced data and variants of known inheritance class, we build an inheritance model and then a de novo score for variants when parental data are missing. Next, we propose a new Random Draw (RD) model to use this score for gene discovery. Built into an existing inferential framework, RD produces a more powerful gene-based association test and controls the false discovery rate. AVAILABILITY AND IMPLEMENTATION: Codes are available at Github (https://github.com/HaeunM/TADA-RD) and Zenodo (DOI: https://doi.org/10.5281/zenodo.18531769).
REACTOR: REgulon Activity analysis and Comparison Tool for single-cell transcriptOmics Research.
Lindén M, Zúñiga Norman SI, Välikangas T … +4 more , Junttila S, Suomi T, Rytkönen KT, Elo LL
Bioinformatics · 2026 May · PMID 42082398 · Full text
SUMMARY: We introduce REACTOR, a computational tool designed to detect differential activity of transcriptional regulators and their target genes (regulons) in single-cell RNA-sequencing data. It expands the currently av...
SUMMARY: We introduce REACTOR, a computational tool designed to detect differential activity of transcriptional regulators and their target genes (regulons) in single-cell RNA-sequencing data. It expands the currently available framework for regulon analysis by introducing a robust statistical test to detect differential regulon activity between conditions, such as disease versus control, with multiple replicates. By contrasting different conditions, REACTOR enables identification of key condition- and cell type-specific regulons. To demonstrate the use of REACTOR, we illustrate its performance in a publicly available COVID-19 dataset. AVAILABILITY: REACTOR R-package together with an implementation vignette are available at https://www.github.com/elolab/REACTOR.
Cell-o1 : training LLMs to solve single-cell reasoning puzzles with reinforcement learning.
Fang Y, Jin Q, Xiong G … +7 more , Jin B, Zhong X, Ouyang S, Yang Y, Zhang A, Han J, Lu Z
Bioinformatics · 2026 May · PMID 42082388 · Full text
MOTIVATION: Large language models (LLMs) have demonstrated strong general reasoning abilities, but applying them to domain-specific tasks such as analysing single-cell RNA sequencing data remains a challenge. A central t...
MOTIVATION: Large language models (LLMs) have demonstrated strong general reasoning abilities, but applying them to domain-specific tasks such as analysing single-cell RNA sequencing data remains a challenge. A central task in this domain is cell type annotation, which is critical for understanding cellular heterogeneity. Although recent foundation models attempt to automate this process, they typically annotate cells independently, without considering batch-level context or providing explanatory reasoning. To address this limitation, we introduce the CellPuzzles benchmark, which reformulates cell type annotation as a batch-level reasoning task. CellPuzzles spans diverse tissues, diseases, and donor conditions, and requires reasoning across the batch-level cellular context to ensure label uniqueness. RESULTS: We find that off-the-shelf LLMs struggle on this task, with the best baseline (OpenAI o1) achieving only 19.0% batch-level accuracy. To fill this gap, we propose Cell-o1, a 7B LLM trained via supervised fine-tuning on distilled reasoning traces, followed by reinforcement learning with batch-level rewards. Cell-o1 achieves state-of-the-art performance, outperforming OpenAI o1 by over 73% and generalizing well across contexts. Further analysis of training dynamics and reasoning behaviors provides insights into batch-level annotation performance and emergent expert-like reasoning. AVAILABILITY AND IMPLEMENTATION: Code and data are available at https://github.com/ncbi-nlp/cell-o1.
Interpretable deep survival analysis of Alzheimer's disease via metabolic genetic variants.
Goo S, Lee S, Chae JW … +2 more , Jung S, Yun HY
Bioinformatics · 2026 Jun · PMID 42063212 · Full text
BACKGROUND: Alzheimer's disease (AD) is a progressive neurodegenerative disease. Traditional models for estimating AD onset cannot capture nonlinear interactions (epistasis) among the numerous genetic variables that cont...
BACKGROUND: Alzheimer's disease (AD) is a progressive neurodegenerative disease. Traditional models for estimating AD onset cannot capture nonlinear interactions (epistasis) among the numerous genetic variables that contribute to AD risk. METHODS: We developed a feedforward neural network (FFN)-Weibull survival model to predict AD onset using large-scale single-nucleotide polymorphism (SNP) data. We integrated an XAI technique, Shapley additive explanations (SHAP), to address the black-box nature of deep learning, interpret model predictions, and quantify the contribution of each genetic factor to AD. RESULTS: The FFN model achieved a mean concordance index of 0.647, demonstrating an approximately 3.6% improvement over the traditional linear baseline (0.625). The FFN-SHAP model validated established findings, identifying APOE E4 as a primary AD risk factor. APOE E2 strongly protected against AD. Metabolic-disorder-related SNPs had conflicting effects, suggesting gene-environment interactions influence AD onset. CONCLUSIONS: By effectively bypassing the combinatorial explosion of interaction terms, the predictive power of an FFN combined with XAI provides a robust methodological tool for identifying the genetic basis of complex diseases, even in cohorts with limited sample sizes. Our model generated novel testable hypotheses regarding the intricate roles of gene-gene and gene-environment interactions in AD pathogenesis.
nf-core/viralmetagenome: A novel pipeline for untargeted viral genome reconstruction.
Klaps J, Lemey P, Bletsa M … +2 more , nf-core community, Kafetzopoulou LE
Bioinformatics · 2026 May · PMID 42057295 · Full text
MOTIVATION: Reconstructing eukaryotic viral genomes from metagenomic data is challenging due to their extensive diversity and potential genome segmentation. Current approaches often rely on labor-intensive manual curatio...
MOTIVATION: Reconstructing eukaryotic viral genomes from metagenomic data is challenging due to their extensive diversity and potential genome segmentation. Current approaches often rely on labor-intensive manual curation for reference selection and scaffolding, limiting scalability for large studies or rapid outbreak response. We address the critical need for an automated, scalable pipeline for efficient viral metagenomic analysis without manual intervention. RESULTS: We present nf-core/viralmetagenome, a comprehensive Nextflow pipeline for the untargeted reconstruction and variant analysis of eukaryotic DNA and RNA viruses from short-read metagenomic or hybridisation capture enriched samples. The pipeline automates the entire process from read preprocessing to consensus generation, integrating multiple de novo assemblers, automated reference selection, and iterative consensus refinement. It features robust quality control, extensive documentation, and seamless portability via Docker and Singularity. We validated the pipeline on diverse simulated and real datasets, demonstrating its ability to recover high-quality genomes from complex metagenomic samples and resolve co-infections, making it a powerful tool for viral surveillance. AVAILABILITY: nf-core/viralmetagenome is freely available at https://github.com/nf-core/viralmetagenome with comprehensive documentation at https://nf-co.re/viralmetagenome. Archival code repository snapshots are published at zenodo with doi: https://doi.org/10.5281/zenodo.17524074.
Metappuccino: large language model-driven reconstruction of sequence read archive metadata for cancer research.
Hak F, Marchet C, Gautheret D … +1 more , Gallopin M
Bioinformatics · 2026 May · PMID 42057294 · Full text
MOTIVATION: High-throughput RNA sequencing has significantly advanced transcriptomic profiling in oncology. Millions of RNA-seq datasets have accumulated in public databases such as the Sequence Read Archive (SRA). Howev...
MOTIVATION: High-throughput RNA sequencing has significantly advanced transcriptomic profiling in oncology. Millions of RNA-seq datasets have accumulated in public databases such as the Sequence Read Archive (SRA). However, fragmented, ambiguous, or missing metadata can severely limit accurate cohort selection, introduce bias, and delay discoveries. RESULTS: To address these issues, we introduce 'Metappuccino', a hybrid metadata enrichment tool built on Mistral-7B-Instruct and specialized via low-rank adaptation (LoRA). Metappuccino reconstructs 19 metadata classes (e.g. organ, disease, cell type) by combining deterministic extraction/normalization with model-based completion: 4 submission-mandatory fields are read directly from SRA/API records, while the remaining 15 classes are obtained through validated rule-based extraction when explicitly supported by the context and otherwise predicted by the LoRA-specialized model when information is missing or ambiguous. To promote robust, context-aware inference rather than memorization, we designed training and data partitioning to minimize leakage and preserve generalization. When applicable, predicted values are mapped to standardized ontologies to ensure consistent, interoperable annotations. Across our benchmarks, Metappuccino substantially improves accuracy over the base model, matches or exceeds recent larger open-source LLMs, and reduces inference time by up to two-fold relative to these baselines. By enriching under-annotated public RNA-seq records, Metappuccino increases the usability of SRA datasets for large-scale reuse, with applications that extend beyond oncology transcriptomics. AVAILABILITY AND IMPLEMENTATION: Metappuccino source code is available on: github.com/chumphati/Metappuccino. The fine-tuned LLM, MetappuccinoLLModel, is available on: huggingface.co/chumphati/MetappuccinoLLModel. Both repositories are released under Apache-2.0 license.
MuFaDDG: a sequence-based multiscale feature fusion framework for protein stability changes prediction.
Gong J, Ma P, Ren Z … +5 more , Li S, Fu Z, Sun P, Ni M, Bo X
Bioinformatics · 2026 May · PMID 42057285 · Full text
MOTIVATION: Predicting the thermodynamic stability of proteins upon single-point mutations is a pivotal step in both protein engineering and medicine. In the study of predicting protein thermodynamic stability, various c...
MOTIVATION: Predicting the thermodynamic stability of proteins upon single-point mutations is a pivotal step in both protein engineering and medicine. In the study of predicting protein thermodynamic stability, various computational methods, whether they extract features at the local-level or global-level, exhibit their respective advantages and limitations. To leverage the advantages of both features, we developed MuFaDDG, a novel sequence-based method that integrated multiscale feature fusion for improved prediction of protein stability changes (ΔΔG). RESULTS: MuFaDDG achieves comparable performance on the S669 benchmark, demonstrating strong capabilities in stabilizing mutations. Notably, it shows a significant advantage in the ACC metric, with values of 0.75, 0.88, and 0.81 on the direct, reverse, and overall datasets of the CAGI5 Challenge's Frataxin, respectively. Furthermore, our method outperforms leading sequence-based approaches including THPLM, DDGemb, DDGun, and INPS-Seq on protein Myoglobin stability prediction. Additionally, MuFaDDG demonstrates exceptional predictive performance with higher PCC and ACC on the protein ThreeFoil, which is uncurated by FireProtDB and ProThermDB databases. AVAILABILITY AND IMPLEMENTATION: The source code and data are available at https://github.com/PengjiaMa23/MuFaDDG.
FUSE: data-driven functional segmentation of DNA methylation data.
Holmström S, Häkkinen A, Lavikka K … +3 more , Marchi G, Hautaniemi S, Lahtinen A
Bioinformatics · 2026 May · PMID 42057265 · Full text
SUMMARY: DNA methylation (DNAm) of neighbouring CpG sites is highly correlated, making DNAm function in terms of blocks. DNAm patterns and functionality are linked to both chromatin structure of DNA and gene regulation....
SUMMARY: DNA methylation (DNAm) of neighbouring CpG sites is highly correlated, making DNAm function in terms of blocks. DNAm patterns and functionality are linked to both chromatin structure of DNA and gene regulation. Defining biologically meaningful DNA methylation blocks from whole-genome bisulfite sequencing (WGBS) data remains challenging, as most existing methods rely on fixed genomic windows rather than the observed methylation pattern. We present FUSE, a data-driven segmentation method that captures intrinsic methylation segments directly from WGBS data by jointly analyzing multiple samples. FUSE identifies spatially homogeneous methylation blocks shared across the input cohort while allowing different methylation states across samples. Applied to 61 WGBS samples from the ENCODE database, FUSE identified segments which overlap significantly with promoters, enhancers, and repetitive elements. FUSE was able to recover the true segment breakpoints in synthetic data with high sensitivity under increased levels of noise. As such, FUSE facilitates post hoc methylation analyses by aggregating coherent CpG sites into candidate segments for downstream differential methylation testing or other comparative studies. AVAILABILITY AND IMPLEMENTATION: FUSE is implemented as an R-package methFuse, available at https://github.com/holmsusa/methFuse and https://cran.r-project.org/package=methFuse. A GenomeSpy visualization of the data is available at https://csbi.ltdk.helsinki.fi/p/fuse_encode_gs/.
Inferring the qualities of protein-RNA models with graph transformers.
Siciliano AJ, Bao Y, Shrestha B … +1 more , Wang Z
Bioinformatics · 2026 May · PMID 42048142 · Full text
MOTIVATION: Breakthrough advancements in protein tertiary and quaternary structure prediction have accelerated structural bioinformatics research activity and drug development processes. However, many biological mechanis...
MOTIVATION: Breakthrough advancements in protein tertiary and quaternary structure prediction have accelerated structural bioinformatics research activity and drug development processes. However, many biological mechanisms involve more complicated interactions, such as those between amino and nucleic acids. Predicting the structure of protein-RNA complexes is highly relevant and challenging due to data scarcity and experimental difficulties. Understanding and interpreting these interactions can yield crucial insights into various human diseases and biological phenomena. Thus, quality assessment methods that specifically evaluate protein-RNA complex models can provide significant utility in this emerging area of protein-RNA structural bioinformatics research. RESULTS: We propose a novel graph transformer-based approach named complex quality assessment of RNA and protein (CARP) to infer multiple quality perspectives of protein-RNA complex models. For a single protein-RNA complex model, in one shot, CARP simultaneously predicts multiple overall fold, overall interface, and per-protein-RNA interface quality estimates. When evaluated against a non-redundant protein-RNA docking benchmark, our methods demonstrated obvious improved performance compared to almost all of the existing scoring tools, particularly when ordering and selecting the highest quality decoys. Furthermore, CARP consistently selected higher quality models relative to other predictors when tested on CASP16 targets. Specifically, CARP-predicted global interface and global protein-RNA interface qualities were ranked first and second, respectively, based on the selected top-3 models over all ten CASP16 protein-RNA complex targets. CARP also showed a strong ability, compared to both existing tools and AlphaFold3 self-estimates, in selecting high quality AlphaFold3 models. AVAILABILITY AND IMPLEMENTATION: CARP is freely available at github.com/zwang-bioinformatics/CARP/.
ChASM: a statistically rigorous method for the detection of chromosomal aneuploidies in ancient DNA studies.
Rohrlach AB, Tuke J, Prüfer K … +1 more , Haak W
Bioinformatics · 2026 May · PMID 42048124 · Full text
MOTIVATION: How individuals with conditions, disabilities or abnormalities were treated gives us valuable insights into past societies. Chromosomal aneuploidies, the presence of an abnormal number of copies of the chromo...
MOTIVATION: How individuals with conditions, disabilities or abnormalities were treated gives us valuable insights into past societies. Chromosomal aneuploidies, the presence of an abnormal number of copies of the chromosomes, represent the most common large-scale chromosomal abnormalities in human populations. Chromosomal aneuploidies can affect autosomal chromosomes (e.g. Down syndrome) as well as the sex chromosomes (e.g. Klinefelter syndrome), with physical manifestations ranging from mild to severe. While simple to identify genetically, chromosomal aneuploidies are difficult to diagnose from skeletal remains alone, as they present skeletal pathologies consistent with many other conditions. RESULTS: Here we present ChASM (Chromosomal Aneuploidy Screening Methodology), a statistically rigorous Bayesian method for detecting full autosomal and sex chromosomal aneuploidies. The method leverages chromosome-wise read counts and takes into account differences in sequencing methodology, genetic coverage and condition rarity to produce posterior probability estimates for the screening of small and large databases of sequence data. AVAILABILITY AND IMPLEMENTATION: To facilitate the ease of use, ChASM has been implemented in R as the package RChASM. RChASM is available under MIT license on the Comprehensive R Archive Network.
Duplex-Indel: a Snakemake pipeline for somatic Indel calling in Tn5 transposase-based duplex sequencing data.
Dong G, Hilal N, Mallett S … +7 more , Jin B, Mao S, Manam MD, Shao DD, Choudhury S, Huang AY, Lee EA
Bioinformatics · 2026 May · PMID 42046229 · Full text
SUMMARY: Duplex-Indel is a novel Snakemake workflow for detecting somatic small insertions and deletions (Indels) from Tn5 transposase-based duplex sequencing data. Duplex-Indel enhances the accuracy of mutation calling...
SUMMARY: Duplex-Indel is a novel Snakemake workflow for detecting somatic small insertions and deletions (Indels) from Tn5 transposase-based duplex sequencing data. Duplex-Indel enhances the accuracy of mutation calling at the single-molecule level by requiring consensus support from both DNA strands for each somatic Indel, minimizing confounding from technical artifacts. Duplex-Indel extends somatic mutation calling in Tn5 transposase-based duplex sequencing data to include Indels. We have demonstrated the accuracy and robustness of Duplex-Indel using cancer cell lines. AVAILABILITY AND IMPLEMENTATION: Source code and documentation are available under the MIT license on GitHub at https://github.com/ealee-lab/duplex-indel and archived on Zenodo at https://doi.org/10.5281/zenodo.19228799.
MS-ConTab: multi-scale contrastive learning of mutation signatures for Pan-Cancer representation and stratification.
Dou Y, Khadre A, Petreaca RC … +1 more , Golrokh M
Bioinformatics · 2026 May · PMID 42046210 · Full text
MOTIVATION: Understanding pan-cancer level mutational landscape offers critical insights into the molecular mechanisms underlying tumorigenesis. While patient-level machine learning techniques have been widely employed t...
MOTIVATION: Understanding pan-cancer level mutational landscape offers critical insights into the molecular mechanisms underlying tumorigenesis. While patient-level machine learning techniques have been widely employed to identify tumor subtypes, cohort-level clustering-where entire cancer types are grouped based on shared molecular features-has largely relied on classical statistical methods. RESULTS: In this study, we introduce a novel unsupervised contrastive learning framework to cluster 43 cancer types based on coding mutation data derived from the COSMIC database. For each cancer type, we construct two complementary mutation signatures: a gene-level profile capturing nucleotide substitution patterns across the most frequently mutated genes, and a chromosome-level profile representing normalized substitution frequencies across chromosomes. These dual views are encoded using TabNet encoders and optimized via a multi-scale contrastive learning objective (NT-Xent loss) to learn unified cancer-type embeddings. We demonstrate that the resulting latent representations yield biologically meaningful clusters of cancer types, aligning with known mutational processes and tissue origins. Our work represents the first application of contrastive learning to cohort-level cancer clustering, offering a scalable and interpretable framework for mutation-driven cancer subtyping. AVAILABILITY AND IMPLEMENTATION: Data and Code are available at: https://github.com/25Nov/MS-ConTab. SUPPLEMENTARY INFORMATION: Supplementary material includes Supplementary Table 1-3 and Supplementary Figure 1, which provide additional data supporting the main results.
Much ado about nothing: modeling amino acid replacement with predicted protein structures.
Buschmann L, Bolz SN, El-Hendi F … +2 more , Malekian N, Schroeder M
Bioinformatics · 2026 May · PMID 42036821 · Full text
MOTIVATION: Substitution matrices like BLOSUM62 model the likelihood of replacement of amino acids in evolution. Substitution matrices are used in protein sequence alignment tasks. Since the introduction of BLOSUM62 over...
MOTIVATION: Substitution matrices like BLOSUM62 model the likelihood of replacement of amino acids in evolution. Substitution matrices are used in protein sequence alignment tasks. Since the introduction of BLOSUM62 over three decades ago, many matrices have been released. Yet, to date, no effort uses large amounts of 3D structures predicted by AlphaFold. RESULTS: Here, we define AFSM, the AlphaFold Substitution Matrix derived from over 20 000 predicted 3D structures following the BLOSUM methodology. We benchmark AFSM against BLOSUM62 and 16 other matrices on five tasks in multiple sequence alignment (MSA) and protein homology search. Our analysis surprisingly reveals that all matrix families perform similarly. Only when there are few sequences in an MSA do BLOSUM62 and AFSM perform better than using no matrix. This suggests that substitution matrices were most beneficial when there was little sequence data. We corroborate this argument by showing that embeddings, which are computed from billions of sequences, perform better than substitution matrices, when sequence data is sparse. Taken together, this suggests that structural data does not improve BLOSUM62. But increased sequence data makes extrapolation with substitution matrices obsolete. Nonetheless, BLOSUM62 continues to capture chemists' intuition on amino acids by providing numerical values implicitly reflecting physicochemical properties, and it remains indispensable for sparse MSAs and direct comparison of two sequences. AVAILABILITY AND IMPLEMENTATION: Data is available from doi.org/10.5281/zenodo.18777546.
Refining sequence-to-expression modelling with chromatin accessibility.
Lapohos O, Fonseca GJ, Emad A
Bioinformatics · 2026 May · PMID 42036810 · Full text
MOTIVATION: Sequence-to-expression models typically do not consider chromatin accessibility, a major factor limiting gene regulation. We hypothesized that supplying accessibility as an input feature would allow a sequenc...
MOTIVATION: Sequence-to-expression models typically do not consider chromatin accessibility, a major factor limiting gene regulation. We hypothesized that supplying accessibility as an input feature would allow a sequence-to-expression model to focus on important open regions of the genome. RESULTS: We found that the performance of such an augmented model was significantly better than that of sequence-only or accessibility-only models with similar architectures. Specifically, its ability to predict the expression of highly variable genes and gene expression in other cell types improved, and higher attribution scores in the input DNA sequences of the augmented model conformed to accessibility, enabling the learning of cell type-specific sequence patterns. Additionally, we show that fine-tuning a pre-trained sequence-only model with both sequence and accessibility can boost performance further and highlight the importance of sequencing depth in sequence-to-expression prediction. AVAILABILITY AND IMPLEMENTATION: Source code is available on GitHub at https://github.com/lapohosorsolya/accessible_seq2exp.
GR2ST: spatial transcriptomics prediction based on graph-enhanced multimodal contrastive learning.
Zhou J, Li S, Han R … +3 more , Wang X, Wang Y, Li J
Bioinformatics · 2026 May · PMID 42036805 · Full text
MOTIVATION: Spatial transcriptomics techniques capture gene expression data and spatial coordinates, while simultaneously correlating them with tissue section images. This advantage makes Spatial transcriptomics data hig...
MOTIVATION: Spatial transcriptomics techniques capture gene expression data and spatial coordinates, while simultaneously correlating them with tissue section images. This advantage makes Spatial transcriptomics data highly valuable for research, such as investigating disease mechanisms and cancer prognosis. However, the extended time and high cost of spatial transcriptomic sequencing currently limit further advancements in this field. The development of numerous deep learning methods aimed at predicting spatial transcriptomics from histology images has advanced significantly. However, these approaches often lack the ability to effectively integrate histology images with spatial transcriptomic data. Here, we propose GR2ST, a deep learning model that learns the underlying connections between image features and gene expression to predict spatial transcriptomics. RESULTS: GR2ST leverages a large pre-trained pathology model to extract high-level histological features. We designed a dual-branch graph architecture, consisting of a dynamic threshold-based functional graph and a radius-constrained spatial graph, to capture complex spot interactions within heterogeneous tissues. The model aligns histology images with gene expression representations through a multimodal contrastive learning framework. It achieves adaptive gene expression generation via a Cell-Type Guided Multi-Branch Regression Head supervised by a context-aware weighting network, which is further integrated with cross-sample retrieval to construct an ensemble prediction. The performance of the model is evaluated on three cancer-related spatial transcriptomics datasets, including cutaneous squamous cell carcinoma and two human breast cancer cohorts, to demonstrate its effectiveness and robustness. AVAILABILITY: https://github.com/zjl1109294570/GR2ST.
Supervised fine-tuning enhances unsupervised learning from 45 million amino acids in TCR and peptide sequences.
Zhou K, Xu K, Lin S … +3 more , Zhai S, Liu H, Yao X
Bioinformatics · 2026 May · PMID 42032806 · Full text
MOTIVATION: T cell receptor (TCR) and peptide interactions (TPI) are one of the most important parts of T cell immunity. Experimental identification of TPI is time-consuming and labor-intensive; therefore, it is necessar...
MOTIVATION: T cell receptor (TCR) and peptide interactions (TPI) are one of the most important parts of T cell immunity. Experimental identification of TPI is time-consuming and labor-intensive; therefore, it is necessary to develop computational prediction method that exploit existing data to predict TPI. RESULTS: We use huge TCR and peptide sequences to pre-train two language models (∼152M parameters), respectively, and integrate them into a sequence-based only prediction framework (i.e. RoBERTcr) with supervised fine-tuning (SFT). Visualization of amino acids embedding from pre-trained language model (PLM) shows biochemical clusters based on different properties, and our PLMs outperform existing protein language models (i.e. ESM and ProtTrans) under the same condition. RoBERTcr achieved higher performance than other state-of-the-art methods based on structures or sequences without dataset bias. The visualization of attention from our framework implies valuable spatial information that residues in TCR contacting peptides are the key to their interaction. AVAILABILITY: RoBERTcr is free available at https://fca_icdb.mpu.edu.mo/robertcr/ and https://doi.org/10.5281/zenodo.18043054.
A cross-attentive multi-task graph learning framework for chemical reaction modeling.
Astero M, Li A, Casiraghi E … +1 more , Rousu J
Bioinformatics · 2026 May · PMID 42026905 · Full text
MOTIVATION: Understanding chemical reactions requires bridging fine-grained molecular edits with broader semantic context. Reaction mechanisms are determined not only by local atom-bond transformations but also by the gl...
MOTIVATION: Understanding chemical reactions requires bridging fine-grained molecular edits with broader semantic context. Reaction mechanisms are determined not only by local atom-bond transformations but also by the global reaction class. However, most existing approaches treat these tasks separately or rely on external atom-mapping tools, introducing noise and limiting end-to-end learnability. We introduce MARCC (Mapping-Assisted Reaction Center and Classification), a multi-task graph neural network that jointly predicts atom mappings, reaction centers, and reaction classes within a unified architecture. RESULTS: MARCC integrates three key innovations: (i) a mapping-guided cross-attention mechanism that aligns reactants and products for local edit detection, (ii) a dual-graph design that explicitly reasons about bond-level transformations, and (iii) pooled product embeddings for global reaction classification. On the USPTO-50K benchmark, MARCC achieves state-of-the-art results when trained with both reactants and products, including 98.2% atom mapping accuracy, 99.1% Top-1 edit localization accuracy, and 97.2% reaction classification accuracy. Even under the products-only setting, MARCC delivers competitive performance comparable to specialized baselines. Ablation studies confirm the value of mapping-guided attention and multi-task supervision, which enhance both predictive accuracy and interpretability. By unifying atom-level alignment, local reactivity, and global classification, MARCC provides a structured and interpretable framework for reaction understanding. Beyond benchmarks, MARCC has the potential to support applications in reaction annotation, template discovery, and mechanism inference; with additional domain-specific modeling and data, it could be extended to biochemical domains such as enzyme-catalyzed transformations and metabolic pathway modeling. AVAILABILITY AND IMPLEMENTATION: The source code and implementation details are available at https://github.com/maryamastero/MARCC and archived at https://doi.org/10.5281/zenodo.18500230.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe