Search for: All records

The field of geometric deep learning has recently had a profound impact on several scientific domains such as protein structure prediction and design, leading to methodological advancements within and outside of the realm of traditional machine learning. Within this spirit, in this work, we introduce GCPNet, a new chirality-aware SE(3)-equivariant graph neural network designed for representation learning of 3D biomolecular graphs. We show that GCPNet, unlike previous representation learning methods for 3D biomolecules, is widely applicable to a variety of invariant or equivariant node-level, edge-level, and graph-level tasks on biomolecular structures while being able to (1) learn important chiral properties of 3D molecules and (2) detect external force fields.

Across four distinct molecular-geometric tasks, we demonstrate that GCPNet’s predictions (1) for protein–ligand binding affinity achieve a statistically significant correlation of 0.608, more than 5%, greater than current state-of-the-art methods; (2) for protein structure ranking achieve statistically significant target-local and dataset-global correlations of 0.616 and 0.871, respectively; (3) for Newtownian many-body systems modeling achieve a task-averaged mean squared error less than 0.01, more than 15% better than current methods; and (4) for molecular chirality recognition achieve a state-of-the-art prediction accuracy of 98.7%, better than any other machine learning method to date.

The source code, data, and instructions to train new models or reproduce our results are freely available at https://github.com/BioinfoMachineLearning/GCPNet.

In metagenomics, the study of environmentally associated microbial communities from their sampled DNA, one of the most fundamental computational tasks is that of determining which genomes from a reference database are present or absent in a given sample metagenome. Existing tools generally return point estimates, with no associated confidence or uncertainty associated with it. This has led to practitioners experiencing difficulty when interpreting the results from these tools, particularly for low-abundance organisms as these often reside in the “noisy tail” of incorrect predictions. Furthermore, few tools account for the fact that reference databases are often incomplete and rarely, if ever, contain exact replicas of genomes present in an environmentally derived metagenome.

We present solutions for these issues by introducing the algorithm YACHT: Yes/No Answers to Community membership via Hypothesis Testing. This approach introduces a statistical framework that accounts for sequence divergence between the reference and sample genomes, in terms of ANI, as well as incomplete sequencing depth, thus providing a hypothesis test for determining the presence or absence of a reference genome in a sample. After introducing our approach, we quantify its statistical power and how this changes with varying parameters. Subsequently, we perform extensive experiments using both simulated and real data to confirm the accuracy and scalability of this approach.

The source code implementing this approach is available via Conda and at https://github.com/KoslickiLab/YACHT. We also provide the code for reproducing experiments at https://github.com/KoslickiLab/YACHT-reproducibles.

Gos is a declarative Python library designed to create interactive multiscale visualizations of genomics and epigenomics data. It provides a consistent and simple interface to the flexible Gosling visualization grammar. Gos hides technical complexities involved with configuring web-based genome browsers and integrates seamlessly within computational notebooks environments to enable new interactive analysis workflows.

Gos is released under the MIT License and available on the Python Package Index (PyPI). The source code is publicly available on GitHub (https://github.com/gosling-lang/gos), and documentation with examples can be found at https://gosling-lang.github.io/gos.

Developing new crop varieties with superior performance is highly important to ensure robust and sustainable global food security. The speed of variety development is limited by long field cycles and advanced generation selections in plant breeding programs. While methods to predict yield from genotype or phenotype data have been proposed, improved performance and integrated models are needed.

We propose a machine learning model that leverages both genotype and phenotype measurements by fusing genetic variants with multiple data sources collected by unmanned aerial systems. We use a deep multiple instance learning framework with an attention mechanism that sheds light on the importance given to each input during prediction, enhancing interpretability. Our model reaches 0.754 ± 0.024 Pearson correlation coefficient when predicting yield in similar environmental conditions; a 34.8% improvement over the genotype-only linear baseline (0.559 ± 0.050). We further predict yield on new lines in an unseen environment using only genotypes, obtaining a prediction accuracy of 0.386 ± 0.010, a 13.5% improvement over the linear baseline. Our multi-modal deep learning architecture efficiently accounts for plant health and environment, distilling the genetic contribution and providing excellent predictions. Yield prediction algorithms leveraging phenotypic observations during training therefore promise to improve breeding programs, ultimately speeding up delivery of improved varieties.

Available at https://github.com/BorgwardtLab/PheGeMIL (code) and https://doi.org/doi:10.5061/dryad.kprr4xh5p (data).

Despite experimental and curation efforts, the extent of enzyme promiscuity on substrates continues to be largely unexplored and under documented. Providing computational tools for the exploration of the enzyme–substrate interaction space can expedite experimentation and benefit applications such as constructing synthesis pathways for novel biomolecules, identifying products of metabolism on ingested compounds, and elucidating xenobiotic metabolism. Recommender systems (RS), which are currently unexplored for the enzyme–substrate interaction prediction problem, can be utilized to provide enzyme recommendations for substrates, and vice versa. The performance of Collaborative-Filtering (CF) RSs; however, hinges on the quality of embedding vectors of users and items (enzymes and substrates in our case). Importantly, enhancing CF embeddings with heterogeneous auxiliary data, specially relational data (e.g. hierarchical, pairwise or groupings), remains a challenge.

We propose an innovative general RS framework, termed Boost-RS that enhances RS performance by ‘boosting’ embedding vectors through auxiliary data. Specifically, Boost-RS is trained and dynamically tuned on multiple relevant auxiliary learning tasks Boost-RS utilizes contrastive learning tasks to exploit relational data. To show the efficacy of Boost-RS for the enzyme–substrate prediction interaction problem, we apply the Boost-RS framework to several baseline CF models. We show that each of our auxiliary tasks boosts learning of the embedding vectors, and that contrastive learning using Boost-RS outperforms attribute concatenation and multi-label learning. We also show that Boost-RS outperforms similarity-based models. Ablation studies and visualization of learned representations highlight the importance of using contrastive learning on some of the auxiliary data in boosting the embedding vectors.

A Python implementation for Boost-RS is provided at https://github.com/HassounLab/Boost-RS. The enzyme-substrate interaction data is available from the KEGG database (https://www.genome.jp/kegg/).

Protein function prediction, based on the patterns of connection in a protein–protein interaction (or association) network, is perhaps the most studied of the classical, fundamental inference problems for biological networks. A highly successful set of recent approaches use random walk-based low-dimensional embeddings that tend to place functionally similar proteins into coherent spatial regions. However, these approaches lose valuable local graph structure from the network when considering only the embedding. We introduce GLIDER, a method that replaces a protein–protein interaction or association network with a new graph-based similarity network. GLIDER is based on a variant of our previous GLIDE method, which was designed to predict missing links in protein–protein association networks, capturing implicit local and global (i.e. embedding-based) graph properties.

GLIDER outperforms competing methods on the task of predicting GO functional labels in cross-validation on a heterogeneous collection of four human protein–protein association networks derived from the 2016 DREAM Disease Module Identification Challenge, and also on three different protein–protein association networks built from the STRING database. We show that this is due to the strong functional enrichment that is present in the local GLIDER neighborhood in multiple different types of protein–protein association networks. Furthermore, we introduce the GLIDER graph neighborhood as a way for biologists to visualize the local neighborhood of a disease gene. As an application, we look at the local GLIDER neighborhoods of a set of known Parkinson’s Disease GWAS genes, rediscover many genes which have known involvement in Parkinson’s disease pathways, plus suggest some new genes to study.

All code is publicly available and can be accessed here: https://github.com/kap-devkota/GLIDER.

Supplementary data are available at Bioinformatics online.

The R programming language is one of the most widely used programming languages for transforming raw genomic datasets into meaningful biological conclusions through analysis and visualization, which has been largely facilitated by infrastructure and tools developed by the Bioconductor project. However, existing plotting packages rely on relative positioning and sizing of plots, which is often sufficient for exploratory analysis but is poorly suited for the creation of publication-quality multi-panel images inherent to scientific manuscript preparation.

We present plotgardener, a coordinate-based genomic data visualization package that offers a new paradigm for multi-plot figure generation in R. Plotgardener allows precise, programmatic control over the placement, esthetics and arrangements of plots while maximizing user experience through fast and memory-efficient data access, support for a wide variety of data and file types, and tight integration with the Bioconductor environment. Plotgardener also allows precise placement and sizing of ggplot2 plots, making it an invaluable tool for R users and data scientists from virtually any discipline.

Package: https://bioconductor.org/packages/plotgardener, Code: https://github.com/PhanstielLab/plotgardener, Documentation: https://phanstiellab.github.io/plotgardener/.

Supplementary data are available at Bioinformatics online.

Membrane proteins are encoded by approximately one fifth of human genes but account for more than half of all US FDA approved drug targets. Thanks to new technological advances, the number of membrane proteins archived in the PDB is growing rapidly. However, automatic identification of membrane proteins or inference of membrane location is not a trivial task.

We present recent improvements to the RCSB Protein Data Bank web portal (RCSB PDB, rcsb.org) that provide a wealth of new membrane protein annotations integrated from four external resources: OPM, PDBTM, MemProtMD and mpstruc. We have substantially enhanced the presentation of data on membrane proteins. The number of membrane proteins with annotations available on rcsb.org was increased by ∼80%. Users can search for these annotations, explore corresponding tree hierarchies, display membrane segments at the 1D amino acid sequence level, and visualize the predicted location of the membrane layer in 3D.

Annotations, search, tree data and visualization are available at our rcsb.org web portal. Membrane visualization is supported by the open-source Mol* viewer (molstar.org and github.com/molstar/molstar).

Supplementary data are available at Bioinformatics online.

This article introduces Vivarium—software born of the idea that it should be as easy as possible for computational biologists to define any imaginable mechanistic model, combine it with existing models and execute them together as an integrated multiscale model. Integrative multiscale modeling confronts the complexity of biology by combining heterogeneous datasets and diverse modeling strategies into unified representations. These integrated models are then run to simulate how the hypothesized mechanisms operate as a whole. But building such models has been a labor-intensive process that requires many contributors, and they are still primarily developed on a case-by-case basis with each project starting anew. New software tools that streamline the integrative modeling effort and facilitate collaboration are therefore essential for future computational biologists.

Vivarium is a software tool for building integrative multiscale models. It provides an interface that makes individual models into modules that can be wired together in large composite models, parallelized across multiple CPUs and run with Vivarium’s discrete-event simulation engine. Vivarium’s utility is demonstrated by building composite models that combine several modeling frameworks: agent-based models, ordinary differential equations, stochastic reaction systems, constraint-based models, solid-body physics and spatial diffusion. This demonstrates just the beginning of what is possible—Vivarium will be able to support future efforts that integrate many more types of models and at many more biological scales.

The specific models, simulation pipelines and notebooks developed for this article are all available at the vivarium-notebooks repository: https://github.com/vivarium-collective/vivarium-notebooks. Vivarium-core is available at https://github.com/vivarium-collective/vivarium-core, and has been released on Python Package Index. The Vivarium Collective (https://vivarium-collective.github.io) is a repository of freely available Vivarium processes and composites, including the processes used in Section 3. Supplementary Materials provide with an extensive methodology section, with several code listings that demonstrate the basic interfaces.

Supplementary data are available at Bioinformatics online.

Network inference algorithms aim to uncover key regulatory interactions governing cellular decision-making, disease progression and therapeutic interventions. Having an accurate blueprint of this regulation is essential for understanding and controlling cell behavior. However, the utility and impact of these approaches are limited because the ways in which various factors shape inference outcomes remain largely unknown.

We identify and systematically evaluate determinants of performance—including network properties, experimental design choices and data processing—by developing new metrics that quantify confidence across algorithms in comparable terms. We conducted a multifactorial analysis that demonstrates how stimulus target, regulatory kinetics, induction and resolution dynamics, and noise differentially impact widely used algorithms in significant and previously unrecognized ways. The results show how even if high-quality data are paired with high-performing algorithms, inferred models are sometimes susceptible to giving misleading conclusions. Lastly, we validate these findings and the utility of the confidence metrics using realistic in silico gene regulatory networks. This new characterization approach provides a way to more rigorously interpret how algorithms infer regulation from biological datasets.

Code is available at http://github.com/bagherilab/networkinference/.

Supplementary data are available at Bioinformatics online.