Large tree-of-life (ToL) scale databases of microbial genomes are powerful resources for exploring genome structure and function from a phylogenomic context. Despite the growing availability of genomic data, large-scale genome annotation is still a challenge, with a considerable fraction of genes remaining as unannotated. Here, to expand the capabilities of database-wide gene function prediction, we used our AnnoTree platform as a corpus for training a Word2vec-based genomic language model (gLM). Machine-learning of genomic grammar patterns across the AnnoTree database revealed functional associations between genes and enabled the inference of function for hypothetical proteins and domains, as we demonstrated by predicting novel type VI secretion proteins. Finally, we implemented a web-server to allow users to interact with the AnnoTree Word2vec model, thus facilitating gene function prediction. Ultimately, our work highlighted the GTDB/AnnoTree database as a powerful training database for gLMs focused on prediction and discovery of microbial gene functions.

The analysis and comparison of gene neighborhoods is a powerful approach for exploring microbial genome structure, function, and evolution. Although numerous tools exist for genome visualization and comparison, genome exploration across large genomic databases or user-generated datasets remains a challenge. Here, we introduce AnnoView, a web server designed for interactive exploration of gene neighborhoods across the bacterial and archaeal tree of life. Our server offers users the ability to identify, compare, and visualize gene neighborhoods of interest from 30 238 bacterial genomes and 1672 archaeal genomes, through integration with the comprehensive Genome Taxonomy Database and AnnoTree databases. Identified gene neighborhoods can be visualized using pre-computed functional annotations from different sources such as KEGG, Pfam and TIGRFAM, or clustered based on similarity. Alternatively, users can upload and explore their own custom genomic datasets in GBK, GFF or CSV format, or use AnnoView as a genome browser for relatively small genomes (e.g. viruses and plasmids). Ultimately, we anticipate that AnnoView will catalyze biological discovery by enabling user-friendly search, comparison, and visualization of genomic data. AnnoView is available at http://annoview.uwaterloo.ca.

MOTIVATION: Statistical detection of co-occurring genes across genomes, known as 'phylogenetic profiling', is a powerful bioinformatic technique for inferring gene-gene functional associations. However, this can be a challenging task given the size and complexity of phylogenomic databases, difficulty in accounting for phylogenetic structure, inconsistencies in genome annotation and substantial computational requirements. RESULTS: We introduce PhyloCorrelate-a computational framework for gene co-occurrence analysis across large phylogenomic datasets. PhyloCorrelate implements a variety of co-occurrence metrics including standard correlation metrics and model-based metrics that account for phylogenetic history. By combining multiple metrics, we developed an optimized score that exhibits a superior ability to link genes with overlapping GO terms and KEGG pathways, enabling gene function prediction. Using genomic and functional annotation data from the Genome Taxonomy Database and AnnoTree, we performed all-by-all comparisons of gene occurrence profiles across the bacterial tree of life, totaling 154 217 052 comparisons for 28 315 genes across 27 372 bacterial genomes. All predictions are available in an online database, which instantaneously returns the top correlated genes for any PFAM, TIGRFAM or KEGG query. In total, PhyloCorrelate detected 29 762 high confidence associations between bacterial gene/protein pairs, and generated functional predictions for 834 DUFs and proteins of unknown function. AVAILABILITYAND IMPLEMENTATION: PhyloCorrelate is available as a web-server at phylocorrelate.uwaterloo.ca as well as an R package for analysis of custom datasets. We anticipate that PhyloCorrelate will be broadly useful as a tool for predicting function and interactions for gene families. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

SUMMARY: A critical step in comparative genomics is the identification of differences in the presence/absence of encoded biochemical pathways among organisms. Our library, Pygenprop, facilitates these comparisons using data from the Genome Properties database. Pygenprop is written in Python and, unlike existing libraries, it is compatible with a variety of tools in the Python data science ecosystem, such as Jupyter Notebooks for interactive analyses and scikit-learn for machine learning. Pygenprop assigns YES, NO, or PARTIAL support for each property based on InterProScan annotations of open reading frames from an organism's genome. The library contains classes for representing the Genome Properties database as a whole and methods for detecting differences in property assignments between organisms. As the Genome Properties database grows, we anticipate widespread adoption of Pygenprop for routine genome analyses and integration within third-party bioinformatics software. AVAILABILITY AND IMPLEMENTATION: Pygenprop is written in Python and is compatible with versions 3.6 or higher. Source code is available under Apache Licence Version 2 at https://github.com/Micromeda/pygenprop. The package can be installed from both PyPi (https://pypi.org/project/pygenprop) and Anaconda (https://anaconda.org/lbergstrand/pygenprop). Documentation is available on Read the Docs (http://pygenprop.rtfd.io/).

Motivation: Spatially clustered mutations within specific regions of protein structure are thought to result from strong positive selection for altered protein functions and are a common feature of oncoproteins in cancer. Although previous studies have used spatial substitution clustering to identify positive selection between pairs of proteins, the ability of this approach to identify functional shifts in protein phylogenies has not been explored. Results: We implemented a previous measure of spatial substitution clustering (the P3D statistic) and extended it to detect spatially clustered substitutions at specific branches of phylogenetic trees. We then applied the analysis to 423 690 phylogenetic branches from 9261 vertebrate protein families, and examined its ability to detect historical shifts in protein function. Our analysis identified 19 607 lineages from 5362 protein families in which substitutions were spatially clustered on protein structures at P3D < 0.01. Spatially clustered substitutions were overrepresented among ligand-binding residues and were significantly enriched among particular protein families and functions including C2H2 transcription factors and protein kinases. A small but significant proportion of branches with spatially clustered substitution also were under positive selection according to the branch-site test. Lastly, exploration of the top-scoring candidates revealed historical substitution events in vertebrate protein families that have generated new functions and protein interactions, including ancient adaptations in SLC7A2, PTEN, and SNAP25 . Ultimately, our work shows that lineage-specific, spatially clustered substitutions are a useful feature for identifying functional shifts in protein families, and reveal new candidates for future experimental study. Availability and Implementation: Source code and predictions for analyses performed in this study are available at: https://github.com/doxeylab/evoclust3d. Contact: [email protected]. Supplementary information: Supplementary data are available at Bioinformatics online.

SUMMARY: mimicMe is a web server for prediction and analysis of host-like proteins (mimics) encoded by microbial pathogens. Users select a host species and any set of pathogen and control proteomes (bacterial, fungal, protozoan or viral) and mimicMe reports host-like proteins that are unique to or enriched among pathogens. Additional server features include visualization of structural similarities between pathogen and host proteins as well as function-enrichment analysis. AVAILABILITY AND IMPLEMENTATION: mimicMe is available at http://mimicme.uwaterloo.ca CONTACT: [email protected].

We apply a 3D motif detection approach to insoluble-carbohydrate binding modules (CBMs), a class of proteins that share a common binding activity but lack a common sequence motif or fold. Key features of insoluble-carbohydrate binding sites were incorporated into a 3D motif detection algorithm, and used in a linear discriminant analysis. Our method effectively discriminated all known type A CBMs from a non-redundant structural dataset and correctly detected known binding sites without any false positives. The algorithm was used to screen a structural database of homology-modeled proteins from tobacco, and the results were experimentally validated using an affinity purification assay and mass spectrometric protein identification. The algorithm correctly predicted CBMs and binding sites not included in the training set, and predicted a previously unknown binding site in the PR-5d protein family. This work highlights the potential of 3D motif detection methods for use in large-scale functional annotation.