Biology subjects

Holmes, I. H.

Publications and source records attributed to Holmes, I. H..

5 recordsLinked to original sources

bioRxiv2026

Insertions, deletions, and exchangeable couplings: a Dirichlet process over TKF92 domains and sites

We investigate the use of Expectation-Maximization (O_SCPLOWEMC_SCPLOW) and variational Bayes for inferring rates and interactions under models of molecular coevolution. We first review O_SCPLOWEMC_SCPLOW theory for continuous-time Markov chains (O_SCPLOWCTMCC_SCPLOWs) and develop it for coevolutionary models, exploiting exchangeability and reversibility symmetries to constrain the parameter dimension. We fit several paired amino-acid coevolutionary models to structural alignments and compare the results to previous work. Our richest model trained on pooled coevolutionary data has explanatory power comparable to CherryMLs Q2 matrix (also trained on pooled data), with one quarter the parameters. However, we observe that aggregation of training data can lead to a form of Simpsons Paradox: a mixture model, whose components are parameter-efficient continuous-time Bayes networks (O_SCPLOWCTBNC_SCPLOWs), resolves signals that wash out when a single model tries to capture everything. These signals include both correlated and anticorrelated hydropathy and volume-packing in coevolving amino-acid pairs, as well as the anticorrelated acid/base compensation that was detected by CherryMLs Q2. We next present a closed-form evidence lower bound (ELBO) for O_SCPLOWCTBNC_SCPLOWs using O_SCPLOWEMC_SCPLOW statistics. Compared to the state of the art in variational modeling of O_SCPLOWCTBNC_SCPLOWs, the Euler-Lagrange equations derived by Cohn et al (JMLR, 2010), our closed-form ELBO is competitive in accuracy, considerably more efficient, simpler, and more stable. We conclude by describing a covariant indel model: a Dirichlet process selecting coevolving sites within TKF92, yielding an Infinite Pair HMM over alignments and structures. This model slightly outperforms TKF92 on a structural-alignment benchmark. Code and data are at https://tkfdp.net/.

bioinformatics↗

bioRxiv2025

Selective State Space Models Outperform Transformers at Predicting RNA-Seq Read Coverage

Transformers are the basis for many state-of-the-art machine learning tools, including those for predicting gene expression data from DNA sequence. The considerable time and cost of training transformer models has motivated development of alternative approaches inspired by ideas from the signal-processing literature, such as state-space models (Mamba), Fourier transforms (Hyena), and wavelet transforms (MultiResNet). To evaluate these methods as potential replacements (or complements) for attention, we developed a software library bilby, implemented using Python and Jax/Flax, providing convolutional, attention, bidirectional Hyena, bidirectional Mamba, and striped-architecture models for supervised multi-task learning in functional genomics. We report a comparison of these architectures, testing several hyperparameters and variations, and reporting performance statistics for the withheld test set as well as downstream SNP classifiers. Relative to models comprising convolution and attention layers (implemented in Python and TensorFlow via the Baskerville library used by the Borzoi software), models comprising convolutional, bidirectional Mamba, and (optionally) attention layers achieve small but consistent improvements in prediction accuracy, for roughly comparable training times and parameter counts, when averaged across all output tracks and data splits (a proportional increase of 3-4% in Pearson R, and 1-2% in r2, with the highest gains achieved when Mamba and attention layers were combined in a striped architecture). In contrast, Hyena (when reimplemented as described in the literature) was not competitive with attention-based models at these tasks, while MultiResNet proved too slow to be practical. The gains in prediction accuracy of the Mamba-based models do not yet translate to significantly improved performance on downstream SNP classification tasks: benchmarks using a GTEx eQTL dataset yield roughly similar results for Mamba- and attention-based classifiers, with attention marginally outperforming Mamba in one metric (a difference of +0.007 in area under ROC) and slightly underperforming by another metric (a difference of -0.006 in Spearman rank correlation). We argue that these results suggest selective state-space models (such as Mamba and Striped Mamba) warrant further exploration for functional genomics tasks. Our code and trained models are publicly available at https://github.com/ihh/bilby.

genomics↗

bioRxiv2022

JBrowse Jupyter: A Python interface to JBrowse 2

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSMotivationC_ST_ABSJBrowse Jupyter is a package that aims to close the gap between Python programming and genomic visualization. Web-based genome browsers are routinely used for publishing and inspecting genome annotations. Historically they have been deployed at the end of bioinformatics pipelines, typically decoupled from the analysis itself. However, emerging technologies such as Jupyter notebooks enable a more rapid iterative cycle of development, analysis and visualization. ResultsWe have developed a package that provides a python interface to JBrowse 2s suite of embeddable components, including the primary Linear Genome View. The package enables users to quickly set up, launch and customize JBrowse views from Jupyter notebooks. In addition, users can share their data via Googles Colab notebooks, providing reproducible interactive views. AvailabilityJBrowse Jupyter is released under the Apache License and is available for download on PyPI. Source code and demos are available on GitHub at https://github.com/GMOD/jbrowse-jupyter. Contactihh@berkeley.edu

bioinformatics↗

bioRxiv2016

BioMake: a GNU Make-compatible utility for declarative workflow management

The Unix \"make\" program is widely used in bioinformatics pipelines, but suffers from problems that limit its application to large analysis datasets. These include reliance on file modification times to determine whether a target is stale, lack of support for parallel execution on clusters, and restricted flexibility to extend the underlying logic program. We present BioMake, a make-like utility that is compatible with most features of GNU Make and adds support for popular cluster-based job-queue engines, MD5 signatures as an alternative to timestamps, and logic programming extensions in Prolog. BioMake is available from https://github.com/evoldoers/biomake under the Creative Commons Attribution 3.0 US license. The only dependency is SWI-Prolog, available from http://www.swi-prolog.org/. Contact: Ian Holmes ihholmes+biomake@gmail.com or Chris Mungall cmungall+biomake@gmail.com.

bioinformatics↗

bioRxiv2016

Historian: accurate reconstruction of ancestral sequences and evolutionary rates

Reconstruction of ancestral sequence histories, and estimation of parameters like indel rates, are improved by using explicit evolutionary models and summing over uncertain alignments. The previous best tool for this purpose (according to simulation benchmarks) was ProtPal, but this tool was too slow for practical use. Historian combines an efficient reimplementation of the ProtPal algorithm with performance-improving heuristics from other alignment tools. Simulation results on fidelity of rate estimation via ancestral reconstruction, along with evaluations on the structurally-informed alignment dataset BAliBase 3.0, recommend Historian over other alignment tools for evolutionary applications. Historian is available at https://github.com/ihh/indelhistorian under the Creative Commons Attribution 3.0 US license. Contact: Ian Holmes ihholmes+historian@gmail.com.

bioinformatics↗