# Improving Multi-Tumor Biomarker Health Check-Up Tests with Machine Learning Algorithms

Source: https://onco.cc/key-papers/paper-onetest-algorithm-taiwan-cancers-2020/  
OnCo record `paper-onetest-algorithm-taiwan-cancers-2020` (Key paper). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

The Taiwanese study behind OneTest's algorithm: in 27,938 people having routine tumour-marker check-ups, a machine-learning score separated those later diagnosed with cancer only moderately well.

## Summary

Wang and colleagues (Chang Gung Memorial Hospital, Linkou, Taiwan, with 20/20 GeneSystems) developed sex-specific cancer screening algorithms and an organ-of-origin algorithm from tumour-marker values, age and sex of 27,938 asymptomatic people screened at a Taiwanese medical centre between May 2001 and April 2015, 342 of whom were later confirmed to have cancer, with external validation at a second centre. The area under the curve was 0.8736 for men and 0.6938 for women in external cross validation. The organ-system algorithm had 81.2% sensitivity and 64.9% specificity for men and 81.7% and 67.5% for women. The authors' disclosure states that 20/20 GeneSystems is commercialising the algorithms under the OneTest brand.

## Fields

- Kind: Key paper
- Last checked: 2026-11-03
- Journal: Cancers
- Year: 2020
- DOI: 10.3390/cancers12061442
- Authors: Wang HY, Chen CH, Shi S, Chung CR, et al.
- Findings: Area under the curve in external validation: 0.87 for men, 0.69 for women.; Organ-system algorithm: about 81% sensitivity at 65 to 68% specificity.
- What it means: An area under the curve of 0.69 for women is weak discrimination; the paper reports risk groups rather than a positive predictive value, so it does not show how many positive OneTest results are cancer.
- Caveats: Retrospective real-world data from health check-ups, with only 342 cancers.; Authors are affiliated with or own stock in 20/20 GeneSystems.

## Sources

- Cancers 2020: https://doi.org/10.3390/cancers12061442
- PubMed: https://pubmed.ncbi.nlm.nih.gov/32492934/
- Wu et al., Computers in Biology and Medicine 2022 (LSTM model on 163,174 people): https://doi.org/10.1016/j.compbiomed.2022.105362

## Connected records

- fronts: [Early Detection & Screening](https://onco.cc/fronts/early-detection/)
- technologies: [Multi-cancer early detection (MCED)](https://onco.cc/technologies/mced/), [Proteomics & phosphoproteomics](https://onco.cc/technologies/proteomics/)
- drugs: [OneTest for Cancer](https://onco.cc/drugs/onetest/)
- companies: [20/20 BioLabs](https://onco.cc/companies/2020-biolabs/)
- institutions: [Chang Gung Memorial Hospital](https://onco.cc/institutions/chang-gung-memorial-hospital/)
- journals: [Cancers](https://onco.cc/journals/cancers-mdpi/)

---
JSON: https://onco.cc/api/v1/entities/paper-onetest-algorithm-taiwan-cancers-2020.json