NetMoST
Summary¶
NetMoST (Network-based Machine learning approach for SubTyping) partitions patients into disease biotypes directly from individual-level polygenic SNP allele biomarker (PHB) profiles, rather than from a single aggregate PRS. It was demonstrated on schizophrenia, where it produced three genetically distinct biotypes later validated against independent neuroimaging and cognitive data.
Method¶
NetMoST builds a patient-similarity network from each individual's profile across many polygenic haplotype biomarkers (PHBs) — allele-count-based genetic features distilled from GWAS-associated loci — and applies network-based clustering to partition patients into biotypes, in contrast to methods that cluster variants into pathways before scoring individuals.[1]
Application: Schizophrenia Biotyping¶
Applied to 141 schizophrenia patients, NetMoST identified three biotypes with distinct dominant genetic signatures:[1]
- Biotype 1 (36.9% of patients): enriched for neurodevelopmental-process genes.
- Biotype 2 (28.4%): enriched for immune-related genetic variation.
- Biotype 3 (34.7%): enriched for genes involved in calcium and neurotransmitter transport.
The three biotypes were independently validated against neuroimaging and cognitive-assessment data collected on the same patients, each showing distinct regional brain activation patterns and cognitive profiles consistent with its dominant genetic signature.[1]
See Also¶
- Individual-First Polygenic Risk Score Clustering — the broader family of approaches (cluster people by their profile across partitioned/component genetic scores) NetMoST belongs to
Citations¶
[1] Wei, X., Dong, S., Su, Z., Tang, L., Zhao, P., Pan, C., Wang, F., Tang, Y., Zhang, W., & Zhang, X. (2023). NetMoST: A network-based machine learning approach for subtyping schizophrenia using polygenic SNP allele biomarkers. arXiv preprint arXiv:2302.00104 (also indexed as PMC9915719). Source: wei2023-netmost.md. Supports: method description, biotype composition, and neuroimaging/cognitive validation above. Location: Full text — Methods and Results.
Note on sourcing: the version consulted is an arXiv preprint (v2, 2023-03-10), not a peer-reviewed journal article; confidence is set to
mediumaccordingly.