# Prioritizing cancer therapeutic genes using BioRank: A biologically-informed PageRank framework

## TL;DR

BioRank is a new method that helps scientists identify important cancer genes by combining biological data with a ranking algorithm.

## Contribution

BioRank integrates biological insights into a PageRank framework to improve therapeutic gene prioritization.

## Key findings

- BioRank outperforms existing methods in identifying known cancer targets from OncoKB.
- It achieves higher Recall@ and nDCG@ metrics across seven cancer datasets.
- BioRank identifies both known and under-explored therapeutic gene candidates.

## Abstract

The identification of therapeutic target genes constitutes a critical yet challenging aspect of cancer research, primarily due to the inherent complexities of biological systems and the heterogeneity of molecular data. This study introduces BioRank, an innovative gene prioritization methodology that extends the traditional PageRank algorithm by integrating biological insights through a custom-designed vector. This vector synthesizes differential gene expression, functional annotations (derived from GO, KEGG, and Reactome), and coexpression similarity to achieve a classification of enhanced biological significance. BioRank was validated using RNA sequencing data from The Cancer Genome Atlas (TCGA), alongside protein–protein interaction networks from HIPPIE across seven cancer datasets. Experimental results illustrate that BioRank effectively facilitates the identification and prioritization of therapeutic target genes. Comparative analysis with previous methodologies indicates that BioRank achieves superior predictive performance concerning both the number of target genes in OncoKB, as well as Recall@ and nDCG@ metrics. BioRank operates as a research instrument designed for hypothesis generation, prioritizing candidate therapeutic target genes based on a specified cancer type and standard molecular/network inputs. This empowers researchers to prioritize genes for subsequent biological validation, such as functional assays, while simultaneously retrieving known targets and identifying under-explored candidates.