GraphPath - A graph attention model for molecular stratification with interpretability based on the pathway–pathway interaction network - Abstract

Sep 21, 2026

Cancer molecular heterogeneity is central to personalized therapy, yet accurate clinical prediction also depends on understanding biological mechanisms at both molecular and pathway levels. GraphPath introduces a knowledge-driven graph neural network using multi-head self-attention to model the pathway–pathway interaction network. Trained on multi-omics prostate cancer data, it improves classification over P-NET and other baselines, validates on external cohorts, and supports interpretable pathway target identification.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Abstract

Motivation

Results

Availability and implementation

Introduction

Systems biology GraphPath: a graph attention model for molecular stratification with interpretability based on the pathway–pathway interaction network Teng Ma 1 and Jianxin Wang 1, 􀀃 1 Hunan Provincial Key Lab on Bioinformatics, School of Computer Science and Engineering, Central South University, Changsha 41083, Hunan, China 􀀃 Corresponding author. Hunan Provincial Key Lab on Bioinformatics, School of Computer Science and Engineering, Central South University, Computing Building 303, South LuShan Road 932, Changsha 41083, Hunan, China. E-mail: [jxwang@mail.csu.edu.cn](jxwang@mail.csu.edu.cn) (J.W. ) Associate Editor: Pier Luigi Martelli Abstract Motivation: Studying the molecular heterogeneity of cancer is essential for achieving personalized therapy. At the same time, understanding the biological processes that drive cancer development can lead to the identification of valuable therapeutic targets. Therefore, achieving accurate and interpretable clinical predictions requires paramount attention to thoroughly characterizing patients at both the molecular and biological pathway levels. Results: Here, we present GraphPath, a biological knowledge-driven graph neural network with multi-head self-attention mechanism that implements the pathway–pathway interaction network. We train GraphPath to classify the cancer status of patients with prostate cancer based on their multi-omics profiling. Experiment results show that our method outperforms P-NET and other baseline methods. Besides, two external cohorts are used to validate that the model can be generalized to unseen samples with adequate predictive performance. We reduce the dimensionality of latent pathway embeddings and visualize corresponding classes to further demonstrate the optimal performance of the model. Additionally, since GraphPath’s predictions are interpretable, we identify target cancer-associated pathways that significantly contribute to the model’s predictions. Such a robust and interpretable model has the potential to greatly enhance our understanding of cancer’s biological mechanisms and accelerate the development of targeted therapies. Availability and implementation: [https://github.com/amazingma/GraphPath](https://github.com/amazingma/GraphPath). 1 Introduction Malignant tumors, commonly referred to as cancer, are a group of diseases in which cells divide continuously and excessively, leading to a major global public health problem (Fitzmaurice et al. 2018, Sung et al. 2021, Xia et al. 2022). Molecular heterogeneity is one of the most important concerns in cancer, meaning that during the growth and proliferation of the tumor cell, the daughter cells exhibit differences at the molecular level, resulting in variations in tumor growth rate, invasion and metastasis patterns, drug sensitivity, and other aspects (DagogoJack and Shaw 2018, Sharma et al. 2019, Jiao et al. 2020, Vitale et al. 2021, Shi et al. 2023). These heterogeneities within the same type of cancer give rise to varying treatment responsesand overall prognosis among patients, posing significant challenges in cancer diagnosis and management (Abeshouse et al. 2015, Ateeq et al. 2016, Punt et al. 2017, Buikhuisen et al. 2020). Recent advances in genomics, proteomics, and molecular pathology have led to the discovery of candidate biomarkers that hold potential clinical value for cancer subtype classification, staging, and prognosis (McGuire et al. 2017, Audenet et al. 2018, Zhao et al. 2019, 2020). For example, numerous studies in recent years have demonstrated that molecular subtypes can be identified in most human cancers and that different subtypes are significantly correlated with different therapeutic responses and clinical outcomes (McConkey and Choi 2018, Torres and Grippo 2018, Bozic and Wu 2020, Turajlic et al. 2020, Wolf et al. 2022). The application of molecular profiling technologies creates opportunities for personalized medicine. Projects such as The Cancer Genome

FAQs

What clinical goal does GraphPath aim to support?

It targets accurate and interpretable clinical predictions by characterizing patients at molecular and biological pathway levels.

How does GraphPath model biological pathways?

It uses a knowledge-driven graph neural network with multi-head self-attention to implement a pathway–pathway interaction network.

How was GraphPath evaluated and validated?

It was trained to classify prostate cancer status using multi-omics profiling, outperformed multiple baselines, and was validated on two external cohorts for generalization.

Sep 21, 2026

Uploaded by Ava Thompson