Biology subjects

Burkholz, R.

Publications and source records attributed to Burkholz, R..

2 recordsLinked to original sources

bioRxiv2021

Using Graph Convolutional Neural Networks to Learn a Representation for Glycans

As the only nonlinear and most diverse biological sequence, glycans offer substantial challenges for computational biology. These complex carbohydrates participate in nearly all biological processes - from protein folding to the cellular entry of viruses - yet are still not well understood. There are few computational methods to link glycan sequences to functions and those that do exist do not take full advantage of all the available information of glycans. SweetNet is a graph convolutional neural network model that uses graph representation learning to facilitate a computational understanding of glycobiology. SweetNet explicitly incorporates the nonlinear nature of glycans and establishes a framework to map any glycan sequence to a representation. We show that SweetNet outperforms other computational methods in predicting glycan properties on all reported tasks. More importantly, we show that glycan representations, learned by SweetNet, are predictive of organismal phenotypic and environmental properties. Finally, we present a new application for glycan-focused machine learning, the prediction of viral glycan-binding, that can be used to discover new viral receptors and monitor rapidly mutating viruses.

bioinformatics↗

bioRxiv2020

Gene Regulatory Network Inference as Relaxed Graph Matching

Gene regulatory network inference is instrumental to the discovery of genetic mechanisms driving diverse diseases, including cancer. Here, we present a theoretical framework for PANDA, an established method for gene regulatory network inference. PANDA is based on iterative message passing updates that resemble the gradient descent of an optimization problem, OTTER, which can be interpreted as relaxed inexact graph matching between a gene-gene co-expression and a protein-protein interaction matrix. The solutions of OTTER can be derived explicitly and inspire an alternative spectral algorithm, for which we can provide network recovery guarantees. We compare different solution approaches of OTTER to other inference methods using three biological data sets, which we make publicly available to offer a new application venue for relaxed graph matching in gene regulatory network inference. We find that using modern gradient descent methods with superior convergence properties solving OTTER outperforms state-of-the-art gene regulatory network inference methods in predicting binding of transcription factors to regulatory regions.

bioinformatics↗