GenePlexus: a web-server for gene discovery using network-based machine learning

Abstract

Biomedical researchers take advantage of high-throughput, high-coverage technologies to routinely generate sets of genes of interest across a wide range of biological conditions. Although these technologies have directly shed light on the molecular underpinnings of various biological processes and diseases, the list of genes from any individual experiment is often noisy and incomplete. Additionally, interpreting these lists of genes can be challenging in terms of how they are related to each other and to other genes in the genome. In this work, we present GenePlexus (https://www.geneplexus.net/), a web-server that allows a researcher to utilize a powerful, network-based machine learning method to gain insights into their gene set of interest and additional functionally similar genes. Once a user uploads their own set of human genes and chooses between a number of different human network representations, GenePlexus provides predictions of how associated every gene in the network is to the input set. The web-server also provides interpretability through network visualization and comparison to other machine learning models trained on thousands of known process/pathway and disease gene sets. GenePlexus is free and open to all users without the need for registration.

Open Biomedical Network Benchmark A Python Toolkit for Benchmarking Datasets with Biomedical Networks

“…Focusing on the optimal model for one or a few tasks of interest is utterly essential. This is because, in practice, experimental biologists often come with only one or a few gene sets and want to either obtain new related genes or reprioritize them based on their relevance to the whole gene set [61]. Therefore, understanding the characteristics of the task that for which a particular model works well over others is the key to making better architectural decisions for a new task of interest and, ultimately, designing new specialized GNN for network biology.…”

“…For example, although Adj-LogReg performs the best overall on the HumanNet-DisGeNET dataset, GCN achieves significantly better predictions (t-test p-value < 0.01) for a handful of tasks (Appendix S3), such as type 2 diabetes mellitus (DOID:9352) and dermatitis (DOID:2723). In addition to overall performance, focusing on the best model for one or a few tasks of interest is more important because, in practice, experimental biologists often come with only one (or a few) gene set(s) and want to either (1) obtain new related genes or (2) reprioritize the set of genes based on their relevance to the whole gene set (Mancuso, Bills, et al, 2022). Understanding the characteristics of the task where a particular model works well is the key to make better architectural decision given a specific task of interest, and furthermore, designing new specialized GNN architectures for network biology.…”

Joint representation of molecular networks from multiple species improves gene classification

“…LogisticRegression . For a more detailed explanation of how GenePlexus uses the SL classifier see [ 4 , 40 , 41 ]. Previous works in our group have extensively benchmarked the GenePlexus method using different shallow learning models [ 4 ] as well as graph neural networks [ 42 ].…”

“…Here, we used logistic regression with L2 regularization with a strength of 1 as implemented by the scikit-learn [39] function sklearn.linear_model.LogisticRegression. For a more detailed explanation of how GenePlexus uses the SL classifier see [4,40,41]. Previous works in our group have extensively benchmarked the GenePlexus method using different shallow learning models [4] as well as graph neural networks [42].…”