1483-P: Identifying Conditional Dependence of Proteins Regulating Islet Biology with Machine Learning
Bibliographic record
Abstract
Introduction & Objective: High-risk single-nucleotide polymorphisms (SNPs) account for only ~20% of type 2 diabetes heritability, implying islet genes greatly influence one another to alter diabetes risk. Correlative analysis does not reveal conditional dependencies between key players regulating islet function whereas machine learning (ML) may do so. Methods: Using data from 374 genetically diverse mice, we derived models predicting islet function from protein abundance with gradient-boosted decision tree algorithms. Some mice (70%) were used to train the models and the rest were used to validate them. We also applied our models to similar data from other mouse studies and to data from humans to see if the models’ accuracies replicated. We then determined if the models included any known islet regulators, proteins with human orthologues possessing glycemia-related SNPs, and enriched for islet-relevant functional pathways. Finally, using the proteins’ predictive influences as meta-traits, we ran quantitative trait locus (QTL) scans to identify genomic regions altering the influence of proteins on islet function. Results: Our analysis revealed < 150 of the ~ 5000 detected proteins sufficiently modeled any one functional trait, with > 90% of a model’s prediction influenced by < 50 proteins. Some models predicted as well or better in the other mouse sets (R > 0.6) and reasonably well in human data (R > 0.5). ML analysis correctly identified the direction of effect for candidates (e.g. DPP8) where prior studies using correlation alone failed. Models for islet traits enriched for pathways potentially relevant for islet function and several unstudied proteins (e.g. CRELD2) present in multiple models have SNPs for glycemia-related traits. Finally, QTL scans revealed genomic regions that may alter the influence of key proteins on islet function and are far from those proteins’ genomic positions. Conclusions: ML can be used to identify conditional dependencies regulating islet function. Disclosure C. Emfinger: None. E. Freiberger: Employee; AbbVie Inc. M. Rabaglia: None. J. Kolic: None. S. Simonett: Employee; Exact Sciences. M. Shortreed: None. L. Clark: None. D. Stapleton: None. K. Schueler: None. K. Mitok: None. T.R. Price: None. J.J. Coon: Research Support; Agilent, Eli Lilly and Company, Merck & Co., Inc. L. Smith: None. M.J. Merrins: None. J.D. Johnson: None. M. Keller: None. A.D. Attie: None. Funding American Diabetes Association (7-21-PDF-157); National Institutes of Health (R01-DK101573-06); National Institutes of Health (GM070683); National Institutes of Health (P41-GM108538); National Institutes of Health (R35GM126914); National Institutes of Health (R35-GM118110); National Institutes of Health (T32-HL007936); National Institutes of Health (R01-DK113103); National Institutes of Health (R01DK127637); Canadian Institutes for Health Research (CIHR) operating grant 168857 CIHR Team Grant (ASD-179092/5-SRA-2021-1149-S-B); CIHR-JDRF Team (ASD-173663/5-SRA-2020-1059-S-B); CIHR Banting fellowship United States Department of Veterans Affairs Biomedical Laboratory Research and Development Service (I01BX005113)
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".