Genome-wide analyses of variance in blood cell phenotypes provide new insights into complex trait biology and prediction

Abstract

Blood cell phenotypes are routinely tested in healthcare to inform clinical decisions. Genetic variants influencing mean blood cell phenotypes have been used to understand disease aetiology and improve prediction; however, additional information may be captured by genetic effects on observed variance. Here, we mapped variance quantitative trait loci (vQTL), i.e. genetic loci associated with trait variance, for 29 blood cell phenotypes from the UK Biobank (N ~ 408,111). We discovered 176 independent blood cell vQTLs, of which 147 were not found by additive QTL mapping. vQTLs displayed on average 1.8-fold stronger negative selection than additive QTL, highlighting that selection acts to reduce extreme blood cell phenotypes. Variance polygenic scores (vPGSs) were constructed to stratify individuals in the INTERVAL cohort (N ~ 40,466), where the genetically most variable individuals had increased conventional PGS accuracy (by ~19%) relative to the genetically least variable individuals. Genetic prediction of blood cell traits improved by ~10% on average combining PGS with vPGS. Using Mendelian randomisation and vPGS association analyses, we found that alcohol consumption significantly increased blood cell trait variances highlighting the utility of blood cell vQTLs and vPGSs to provide novel insight into phenotype aetiology as well as improve prediction.

Fig. 1. vQTLs for 29 blood cell traits and their comparison with additive QTLs.

a Miami plot showing the best (smallest nominal p value, Levene’s test, see methods) vQTL across 29 blood cell traits (top plot) and the corresponding best additive QTLs (bottom plot). Red dots are genome-wide significant independent vQTLs. b Example of pleiotropic effects of the C allele of rs10803164 for the long non-coding RNA LINC02768 on blood cell trait variance. Blue indicates the effect on trait variance had p < 4.6 × 10⁻⁹ (nominal study-wide GWAS significance, Levene’s test, see “Methods” and Data Availability). c Genetic correlation (LDSC) between blood cell trait variance and trait level. Blue indicates the correlation had multi-testing adjusted p < 0.05 (Supplementary Data 6). d Selection coefficient estimated by BayesS for trait variance and level. All analyses used UK Biobank data with sample size ~N ~ 408,111. In panels (b–d), data are presented as mean values ± SEM. Full names of blood cell traits can be found in Supplementary Data 1.

Fig. 2. Relationships between alcohol consumption and blood cell trait variances.

a LocuzZoom plot of variance QTL mapping for platelet crit (pct) variance at ALDH2 gene (Levene’s test, see “Methods”); (b) Mendelian randomization (MR) of alcohol consumption on variance of blood cell traits using GSMR, MR-PRESSO (presso) and weighted-median (wm). Diamonds: significant in 3 methods. c Effects of MR of alcohol consumption on variance of corpuscular haemoglobin concentration (mscv); d Effects of MR of alcohol consumption on variance of corpuscular volume variance (mcv). Dashed fitted lines indicate the coefficient of Mendelian Randomisation (b_xy = 0.07, se_xy = 0.019 for mscv and b_xy = 0.064, se_xy = 0.0188 for mcv). In panels (c, d), multi-test adjusted p values are shown. All analyses used UK Biobank data with sample size ~N ~ 408,111. In panels (b–d), data are presented as mean values ± SEM. Full names of blood cell traits can be found in Supplementary Data 1.

Fig. 3. The variation in the accuracy of PGSs for 27 blood cell traits (Pearson correlation) between the top and bottom vPGS groups.

a Accuracy of PGS in the top vPGS group (more variable group, grey colour) and the difference (orange) of PGS between the top vPGS group (most variable group) and the bottom vPGS group (less variable group). #: count; % percentage; vol: volume; conc: concentration. b Difference of accuracy of PGS between the top and bottom vPGS groups across 27 blood cell traits. ****p (2-side test) <0.0001. For each box, the minimum is the lowest point, the maximum is the highest point, whiskers are maxima 1.5 times of interquartile range, the bottom bound, middle line and top bound of the box are the 25th percentile, median and the 75th percentile, respectively.

Fig. 4. Effects of interaction between PGS and vPGS on blood cell traits.

a Effects of interaction across 27 traits in INTERVAL (Supplementary Data 12). The vertical dashed line indicates the z-score test statistic value = 1.96 which equals nominal p value = 0.05 and bars with z-score value > 1.96 (nominal 2-sided p < 0.05) are in orange colour. #: count; % percentage; vol: volume; conc: concentration. b, c Examples of visualised effects of interaction for eosinophil percentage of white cells (eo_p) and neutrophil count (neut). Data are presented as mean values ± SEM.

Fig. 5. The difference in the variance explained (R²) between PGS models with or without vPGS.

Each bar represents the relative increase in R² (model goodness of fit) for the blood cell trait when the PGS model added vPGS. In the left panel, the single-trait vPGS was added to PGS. In the right panel, multi-trait vPGS was added to PGS. #: count; % percentage; vol: volume; conc: concentration. *p < 0.05; **p < 0.01; ***p < 0.001 and ****p < 0.0001. nominal 2-sided p values were estimated by comparing models with and without vPGS using r2redux.

Fig. 6. Association between BMI, age, alcohol drinking and smoking and individuals to be genetically variable across blood cell traits in INTERVAL.

a An overall Z score test estimate across 27 blood cell traits. b Z score test estimates for mean corpuscular volume (mcv), neutrophil percentage of white cells (neut_p) and red blood cell count (rbc) which were significant Mendelian Randomisation analyses. Z score = beta (effects) / se (standard error). * (nominal 2-sided): p < 0.05; **: p < 0.01; ***: p < 0.001 and **** p < 0.0001.