OncoMatch - Optimizing Oncology Combination Therapy Prediction Through Genomic, Structural, and Network Analysis

Available at: https://oncomatchapp-precision-medicine.streamlit.app

OncoMatch was created by Ben Busby, Lisa Boatner, Shelby Kroeger, Stan Gizinski, Manasi Ghogare, and Stu Angus.

About: https://osf.io/preprints/biohackrxiv/vtq5f_v1

Advances in precision medicine are reshaping cancer treatment by tailoring therapies to a patient’s specific genetic profile. Despite this, matching cancer mutations to effective drugs remains a complex task due to variability in mutations across cancer types and limited tools for practical clinical application. In this project, initially developed during the BioIT Hackathon 2025, we created OncoMatch—an open-data-powered web application designed to bridge this gap by integrating genomic, transcriptomic, proteomic, and drug-target interaction data to support therapy selection. Building on prior work in colorectal cancer, we expanded our scope to include bladder, ovarian, and non-small cell lung cancer (NSCLC), using the COSMIC and DrugCentral databases to identify relevant gene mutations and therapeutics. We developed two novel scoring systems—the Cancer Precision Score (CPS) and Gene Precision Score (GPS)—to evaluate drug specificity and potential effectiveness. Using data from DrugCentral, LINCS L1000, and DeepCoverMOA, we created a unified bioactivity dataset for over 4,000 drugs, including measures such as IC50 and Kd values. The OncoMatch platform features interactive tools to visualize drug bioactivity, assess multiomic and structural similarity among compounds, and explore potential drug combinations. Users can query drugs by cancer type and gene mutation, generating insights into the most promising therapies and alternatives. Our open source approach not only democratizes access to high quality bioinformatics tools but also encourages data driven, personalized cancer care. Future directions include refining subtype level predictions and improving the platform’s utility for combinatorial therapy planning. We have developed a streamlit app to make it easy to access this data. That app can be found at https://oncomatchapp-precision-medicine.streamlit.app.

What's the problem?

Despite advancements in precision medicine, identifying effective cancer treatments remains a challenge due to the complexity of linking genetic data to drug pathways. Prior research - has demonstrated a method for identifying potential colorectal cancer drugs by integrating genetic data with drug effectiveness scoring. However, this approach has not been widely applied to other cancers or optimized for clinical decision-making.

How are we solving the problem?

We aim to solve this problem by expanding the existing methodology to other cancers with similar classification systems, including bladder, ovarian, and small-cell lung cancer, thereby broadening its impact. Additionally, we plan to enhance diagnostics by leveraging Drug Central to map current treatments to genetic cancer data, and developing an evidence-based scoring system to assess the effectiveness of drug combinations in clinical settings. By advancing these efforts, we seek to improve precision medicine by making cancer treatment more personalized and data-driven.

Workflow for OncoMATCH

OncoMatch is a web app that matches the cancer and gene mutation type to the most effective drug therapy based on the activity score of the drug against the specific gene target. We utilize the COSMIC database - to identify the gene mutations for specific cancer types and the Drug Central database - to identify the drug therapies and the specific genes they target.

Datasets for OncoMATCH

Cosmic_CancerGeneCensus_v99_GRCh38.tsv

This data comes from a Gene Census located in the Cosmic database. Columns: GENE_SYMBOL, NAME, COSMIC_GENE_ID, CHROMOSOME, GENOME_START, GENOME_STOP, CHR_BAND, SOMATIC, GERMLINE, TUMOUR_TYPES_SOMATIC, TUMOUR_TYPES_GERMLINE, CANCER_SYNDROME, TISSUE_TYPE, MOLECULAR_GENETICS, ROLE_IN_CANCER, MUTATION_TYPES, TRANSLOCATION_PARTNER, OTHER_GERMLINE_MUT, OTHER_SYNDROME, TIER, SYNONYMS

EMA_Approved.csv

This data comes from Drug Central, containing a single column of drug names which have been approved by the (European Medicines Agency) EMA

FDA_Approved.csv

This data comes from Drug Central, containing a single column of drug names which have been approved by the (Food and Drug Administration) FDA

PDMA_Approved.csv

This data comes from Drug Central, containing a single column of drug names which have been approved by the (Product Development and Management Association) PDMA

drug.target.interaction.tsv

This data comes from Drug Central, containing many columns including drug name, the gene it targets, and efficacy metrics/values Columns: DRUG_NAME, STRUCT_ID, TARGET_NAME, TARGET_CLASS, ACCESSION, GENE, SWISSPROT, ACT_VALUE, ACT_UNIT, ACT_TYPE, ACT_COMMENT, ACT_SOURCE, RELATION, MOA, MOA_SOURCE, ACT_SOURCE_URL, MOA_SOURCE_URL, ACTION_TYPE, TRL, ORGANISM

structures.smiles.tsv

This data comes from Drug Central, containing many columns including SMILE data and ID Columns: SMILES, InChl, InChlKey, ID, INN, CAS_RN

drug.structure.id.csv

This data comes from Drug Central, containing 2 columns mapping drug name to struct id Columns: DRUG_NAME, STRUCT_ID

drug.target.interaction.fda.cosmic.cancer.type

This data comes from Drug Central, containing many columns including drug name, the gene it targets, efficacy metrics/values, and cancer type Columns: DRUG_NAME, STRUCT_ID, TARGET_NAME, TARGET_CLASS, ACCESSION, GENE, SWISSPROT, ACT_VALUE, ACT_UNIT, ACT_TYPE, ACT_COMMENT, ACT_SOURCE, RELATION, MOA, MOA_SOURCE, ACT_SOURCE_URL, MOA_SOURCE_URL, ACTION_TYPE, TDL, ORGANISM, Driver_Gene, FDA_Approved, EMA_Apprived, PMDA_Approved, Bladder, Colon, NSCLC, Ovarian

drug.target.interaction.scores.tsv

This data comes from Drug Central, containing many columns including drug name gene targeted, and precision score for each cancer type Columns: DRUG_NAME, nGENE, BLADDER_CPS, BLADDER_CGPS, COLON_CPS, COLON_CGPS, NSCLC_CPS, NSCLC_CGPS, OVARIAN_CPS, OVARIAN_CGPS

gene.id.csv

This data comes from Drug Central, containing 2 columns, mapping gene name to cancer type Columns: GENE, CANCER

1_preprocess_raw_data.ipynb

This notebook processes the DrugCentral and Cosmic raw data for visualizations

2_calculate_drug_score.ipynb

This notebook calculates the drug centric scores for drug-cancer type gene precision and drug-gene type precision

3_calculate_drug_structure_similarity.ipynb

This notebook calculates the drug structure similarities based on the Tanimoto Similarity Score

4_deepcoveragemoa.ipynb

This notebook processes the proteomics raw data (DeepCoverageMOA)

Sample Radar Plots for OncoMATCH