Genomic predictors of the maximal O₂ uptake response to standardized exercise training
21-SNP polygenic score for trainability
Background
Follow-up to the HERITAGE Family Study (REF-429). In 473 sedentary white adults from 99 families, the heritability of the VO₂max gain after a standardized 20-week exercise program was estimated at 47%; this study searched for the specific loci behind it.
Participants and Method
n=473 sedentary adults, all white, from 99 HERITAGE families; standardized 20-week training program
GWAS on 324,611 SNPs for association with the VO₂max gain
Single-SNP analysis: 39 SNPs associated with the gain at P < 1.5 × 10⁻⁴
Stepwise multiple regression of the 39 → a panel of 21 SNPs
Protocol
Genome-wide association study: 473 sedentary adults, all whites, from 99 families of the HERITAGE Family Study, each exposed to a standardized 20-wk exercise program (from the abstract).
324,611 SNPs were tested for association with the gain in VO2max: single-SNP analysis, then stepwise multiple regression of the associated SNPs to build a predictor panel.
Selected SNPs were tested for replication in STRRIDE (n = 183), DREW (n = 112) and HERITAGE blacks (n = 247).
Key Findings
The 21-SNP panel accounted for 49% of the variance in VO₂max trainability
≤9 favorable alleles: mean VO₂max gain 221 ml/min; ≥19 favorable alleles: mean gain 604 ml/min — about 2.7× (arithmetic on the abstract's means)
Strongest single association: rs6552828 in the ACSL1 gene (acyl-CoA synthase long-chain member 1), by itself ~6% of the training response
Other strongest predictors were nearest to PRDM1, GRIN3A, KCNH8 and ZIC4
Replication Status (from the abstract)
ZIC4-nearest SNP replicated in STRRIDE (n=183; sedentary, overweight, dyslipidemic adults aged 40-65)
Two SNPs replicated in DREW (n=112; sedentary obese white women): rs1956197 near DAAM1 and rs17117533 near NDN
Two SNPs (rs884736 in CAMTA1; rs17581162 upstream of RGS18) replicated in HERITAGE blacks (n=247)
The authors call for large-scale replication studies; the full 21-SNP panel was not replicated as a whole in any external cohort
Mechanism Insights
No single dominant gene; a polygenic architecture — the top SNP explains ~6%, the panel of 21 explains 49%
Gene functions suggested by the abstract's names (acyl-CoA synthesis; transcription regulation; glutamate receptor; K⁺ channel; a zinc-finger protein) span metabolism and neural signalling — the abstract does not test mechanism
Practical Application
In 473 sedentary adults, a 21-SNP panel explained 49% of the variance in the VO₂max response to 20 weeks of standardized training; people with ≤9 favorable alleles gained 221 ml/min on average, those with ≥19 gained 604 ml/min — but only a few of the SNPs replicated in other cohorts, and the panel as a whole has not — a strong demonstration that trainability is genetic and polygenic, not a usable individual test
Not for individual prescription — a discovery-cohort panel over-fits its own data; the 49% is an upper bound, not what a commercial test would deliver
Useful for understanding why population-level responder distributions are wide
Commercial DTC tests built on panels like this overstate their predictive power (REF-440)
Limitations
Discovery cohort only (all white, sedentary); only selected SNPs replicated in other cohorts
21 SNPs jointly explain 49% — 51% of the variance in trainability unexplained (arithmetic)
Effect requires the polygenic panel, not single-SNP testing
Phenotype = VO₂max only; other adaptation pathways not addressed
How this entry was checked and changed
Why this level
regraded 2026-09-27 from A: discovery GWAS in a single cohort (n=473) with replication of only some SNPs in three other cohorts
Content correction 2026-09-27
"Japanese replication cohort failed; Norwegian cohort partial" -> not in the abstract; replaced with the abstract's replication results (STRRIDE n=183, DREW n=112, HERITAGE blacks n=247, specific SNPs only).
"ACOXL (mitochondrial biogenesis)", "linear dose-response across alleles", "each SNP individually accounts for 1-6%" -> not in the abstract; removed. Gene list replaced with the abstract's PRDM1, GRIN3A, KCNH8, ZIC4.
Added 39 SNPs at P < 1.5 × 10⁻⁴ and the study population (all white, 99 families). Level A -> B+ (single-cohort GWAS); added Tier/Sport lines.
2026-09-28: added Protocol from the abstract (Europe PMC PMID 21183627).
2026-10-01: cross-reference to SYNTH-13 no longer names an archetype — that model was retired as an unvalidated composite.