PathwayGNN

BPP: BPP: A Platform for Automatic Biochemical Pathway Prediction

Authors: Xinhao Yi, Siwei Liu, Shiyang Liang, Yu Wu, Douglas McCloskey*, Zaiqaio Meng*

Contributions

1. We develop BPP, an open-source biochemical pathway analysis platform dedicated to predicting potential links and node attributes in biochemical pathway networks.

2. Based on BPP, we evaluate the performance of four representation learning models on four biochemical pathway datasets. Experimental results suggest that these automated prediction models can achieve reliable performance on link prediction and attribute prediction task.

3. BPP integrates an explainer that provides an interpretation of the prediction results, i.g., offering the contribution of nodes and attributes within the reaction for current prediction result.

4. We verify the effectiveness of BPP by conducting a case study on SARS-CoV-2's invasion process. The results indicate that BPP can successfully identify unseen links within pathways.

Motivation

1. Identifying potential links in biochemical pathway networks is essential for targeting disease markers, discovering drug targets, reconfiguring metabolic networks and addressing gaps in pathways holes in biosynthesis.

2. Traditional experimental methods can impose significant time and labour burdens on researchers, due to a vast number of candidates, consequently, our goal is to enhance the efficiency of pathway studies.

Website

Link to BPP Platform

User Guidance

In this part, we'll introduce how to use our plaform

Customise Biochemical Reaction