Machine learning methodology using a masked neural network for robust genetic risk score calculation from noisy and missing data
Steven Squires, Michael N Weedon, Richard A Oram · 2026
Read Machine learning methodology using a masked neural network for robust genetic risk score calculation from noisy and missing data by Steven Squires, Michael N Weedon, Richard A Oram (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13630461.txt.
Description
Steven Squires, Michael Weedon, and Richard Oram present a 2026 paper on genetic risk scores that survive messy data. Such scores sum many small gene effects to forecast disease risk. Noise or gaps in the input can spoil them. The team built a neural network, a masked multi-layer perceptron, trained on clean scores with noise deliberately mixed in.
A type 1 diabetes score served as the test case. With a tenth of the input corrupted, the plain score's match to the clean version sagged badly. The network's version held a far stronger match. Disease separation also dipped much less under the network. Extra inputs and noisy training made the model sturdier, the authors say. Clinics with imperfect gene data stand to gain.
Catalog record
| CREATOR | Steven Squires, Michael N Weedon, Richard A Oram |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |