Mechanism-aware target discovery

OmniTC

Omics-guided Multi-scale Network Integration for Target Combination prediction, a cancer target-combination prediction framework with generalization ability and mechanism interpretability.

Start Prediction Documentation

Prediction service

Disease-specific target-combination scoring

Submit Gene-symbol target pairs or a CSV/TSV/TXT file, choose a disease, and rank candidates by OmniTC score.

Prediction

Score target pairs

Disease Pathway threshold 0 to 1. Default: 0.50.

Gene-symbol target pairs One pair per line. Separators: comma, tab, semicolon, or spaces. Up to 100 pairs. CSV Upload target-pair file Columns: gene1,gene2. Max file size: 5 MB.

Run pathway analysis

Reactome Local GO BP Optional GO CC Optional GO MF Optional KEGG Optional

For better model co-building, we appreciate your supplemental target combinations. Submit supplement

Results

Ranked combinations

Results will appear here

Scores are returned in descending order with pathway counts when enabled.

Rank	Disease	Gene 1	Gene 2	Score	Label	Shared pathways

Documentation

Using OmniTC

A concise guide for prediction, pathway analysis, result interpretation, and model co-building submissions.

Input format

Use HGNC Gene symbols, one target pair per line, or upload a CSV/TSV/TXT file with gene1 and gene2 columns. Each prediction request accepts up to 100 pairs and file uploads up to 5 MB.

Pathway analysis

The pathway module follows the original filter_and_analyze workflow. Reactome is available locally; GO BP, GO CC, GO MF, and KEGG are optional gseapy-backed categories.

Result columns

Score is the model probability used for descending ranking. Valid indicates the model recognized and scored the submitted target pair.

Supplemental curation

Supplemental target combinations require a reference source. Files are limited to 10 MB, and the local curation folder accepts up to 100 MB total.

Submit supplement

Framework

OmniTC framework

OmniTC models cancer target-combination identification as a matching problem between disease-state representation and target-combination representation.

It constructs multi-omics node features, a heterogeneous gene regulatory network, disease-specific subnetworks, and target-combination representations, then outputs disease-specific target-combination scores through a cross-attention module.

Contact

Lab Location
Laboratory of Molecular Modeling and Design
East China University of Science and Technology

Prof. Yun Tang
ytang234@ecust.edu.cn

Model co-building

Submit supplemental target pairs

Thank you for contributing target-combination evidence. These submissions support future model and website updates.

Disease Contributor

Default reference source Required unless every uploaded row has its own source/reference column. Supplemental entries Format: gene1,gene2,reference. A global reference can fill the third column. REF Upload supplemental file Columns: gene1,gene2,reference_source. Max file size: 10 MB.

Notes