Time-series eQTL Analysis — Dynamic Genetic Regulation of Gene Expression

A standard (static) eQTL study tests whether a SNP is associated with expression level at a single time point or averaged across conditions. A time-series eQTL study tests whether the SNP's effect on expression is different at different time points — the key question is the genotype-by-time interaction. Static eQTLs can be detected with smaller samples; dynamic eQTLs require larger samples but reveal regulatory relationships that are invisible in cross-sectional data.

There is no universal threshold, but as a practical guide you need at least 50–100 genotyped individuals per time point to have reasonable power for detecting interaction effects. Power simulations specific to your effect-size assumptions and number of time points are strongly recommended before committing to data collection. The richer the time series (more time points), the more powerful joint trajectory models become relative to pairwise comparisons.

Standard eQTL tools such as FastQTL and tensorQTL can be adapted for pairwise time-point comparisons. For full dynamic modeling, dedicated frameworks include CSIDE (cell-type-specific inference of differential expression), and custom implementations using linear mixed-effects models in R (lme4, nlme) or Python. Strober et al. (2019) released scripts accompanying their Science paper. The choice depends on whether you treat time as categorical or continuous.

Yes. Time-series eQTL analysis is applicable to any longitudinal study where the same individuals are profiled at multiple time points — aging cohorts, repeated clinical sampling during disease progression, or before-and-after treatment designs. The key requirements are matched genotype and expression data at each time point and sufficient sample size. In vivo studies face additional confounders (lifestyle, medication, cell-type shifts in blood) that must be carefully modeled.

The interaction test (genotype × time) formally distinguishes dynamic from static effects. A significant interaction p-value, after permutation-based FDR correction, indicates that the effect size genuinely differs across time — not just that it is significant at one time point and borderline at another. Replication in an independent cohort or experimental validation (e.g., allele-specific expression at the relevant stage) provides further confidence.