TRV-2026-3021Version 1 · Published

Saved · see the current record

Why this version exists

Certified into the record

The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-3021
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-18T06:02:37.798127Z
status: published
lens: g_space
sector: science
headline: Machine learning-based proteomics analysis identified 23 protein biomarkers for enhanced risk prediction of myocardial infarction in Chinese and Europeans
dek: Background Advances in proteomics assays enable measurement of thousands of plasma proteins, which could facilitate machine learning-based approaches to enhance risk prediction of disease. Methods We measured plasma levels of >9500 proteins among 4000 Chinese adults (1937 myocardial infarction [MI] cases and 2026 subcohort), using both Olink Explore (2923 proteins) and SomaScan (6597 proteins) platforms. Protein-based risk prediction models were constructed, using conventional and machine learning approaches, wi…
gain_title: In a study of 4,000 Chinese adults, a machine-learning score built from 23 blood proteins raised heart attack risk prediction (C-statistic 0.846 to 0.864) beyond conventional risk factors alone.
problem_title: (none)
trace_subject: (none)
gain_reading: In a study of 4,000 Chinese adults, a machine-learning score built from 23 blood proteins raised heart attack risk prediction (C-statistic 0.846 to 0.864) beyond conventional risk factors alone.
gain_evidence: from 0.846 (0.829-0.862) to 0.864 (0.849-0.879)
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers measured more than 9,500 plasma proteins in 4,000 Chinese adults, including 1,937 people who had a myocardial infarction, and used machine learning to pick the proteins that best predicted heart attacks. A score built from 23 selected proteins gave a 3.3-fold difference in 10-year heart attack incidence between the top and bottom 20% risk groups, against 2.5-fold for conventional risk factors and 1.5-fold for polygenic scores.

Adding the protein score to conventional risk factors raised the C-statistic from 0.846 to 0.864 and gave a 72% net reclassification index. In 37,187 Europeans in UK Biobank, adding the same 23 proteins lifted the C-statistic from 0.710 to 0.733, with a 45% net reclassification improvement.

The authors say the results show potential for small-panel proteomics assays to improve disease risk prediction. Such an assay has not yet been built or tested in practice.
limitation: The authors describe only the potential for small panel proteomics assays; no such assay was built or tested, and the UK Biobank validation had just 630 heart attack cases.
tag: Evidence-backed gain
key_points: Plasma levels of more than 9,500 proteins were measured with two platforms: Olink Explore (2,923 proteins) and SomaScan (6,597 proteins). | After adjusting for conventional risk factors, 451 Olink and 530 SomaScan proteins were significantly linked to heart attack risk at a 5% false discovery rate. | Protein models built on Olink, SomaScan or both outperformed conventional risk factors and polygenic scores on discrimination, classification and calibration. | The 23-protein score (11 from Olink, 12 from SomaScan) put 3.6 per 100 people in the top 20% risk group versus 1.1 per 100 in the bottom 20%. | In UK Biobank, 630 heart attack cases among 37,187 Europeans gave a C-statistic gain from 0.710 to 0.733 after adding the 23 proteins.
rundown: The study was built on 1,937 myocardial infarction cases plus a subcohort of 2,026 Chinese adults. Risk models were constructed with both conventional and machine-learning methods, and the Boruta algorithm selected the 23-protein subset. Performance was judged by C-statistic, net reclassification index and calibration.

Conventional risk factors alone separated top and bottom risk groups 2.5-fold (2.5 vs 1.0 per 100), and polygenic scores did so 1.5-fold (1.9 vs 1.3 per 100). The protein score reached 3.3-fold. Combined with conventional risk factors, it lifted the C-statistic by 0.018 in the Chinese cohort.

External validation used 37,187 UK Biobank participants. The baseline C-statistic there was lower, 0.710, and the 23 proteins raised it by 0.023 to 0.733. The authors conclude that machine learning can find smaller protein subsets worth developing into panel assays.
sources:
- peer_reviewed | GigaScience | https://doi.org/10.1093/gigascience/giag102 | 2026-11-17
prev: 0000000000000000000000000000000000000000000000000000000000000000
SHA-256 hash
b542e89057c11a1657ee698c7291f2d7938ac6d8703e7027a8056757a0c51b16
Previous version’s hash
0000000000000000000000000000000000000000000000000000000000000000
Check this record hasn’t been altered
How to check it yourself, without trusting this page

Download any version from /api/record/TRV-2026-3021 and hash its canonical field yourself, for example with shasum -a 256. You should get its content_hash. Each version’s text also ends with prev:and the hash of the version before it (version 1 points to 64 zeros). If anyone had changed a single character of any version after it was saved, those hashes wouldn’t line up.