Genetic prediction of male pattern baldness based on large independent datasets

Abstract

Genetic prediction of male pattern baldness (MPB) is important in science and society. Previous genetic MPB prediction models were limited by sparse marker coverage, small sample size, and/or data dependency in the different analytical steps. Here, we present novel models for genetic prediction of MPB based on a large set of markers and large independent subsample sets drawn among 187,435 European subjects. We selected 117 SNP predictors within 85 distinct loci from a list of 270 previously MPB-associated SNPs in 55,573 males of the UK Biobank Study (UKBB). Based on these 117 SNPs with and without age as additional predictor, we trained, by use of different methods, prediction models in a non-overlapping subset of 104,694 UKBB males and tested them in a non-overlapping subset of 26,177 UKBB males. Estimates of prediction accuracy were similar between methods with AUC ranges of 0.725-0.728 for severe, 0.631-0.635 for moderate, 0.598-0.602 for slight, and 0.708-0.711 for no hair loss with age, and slightly lower without, while prediction of any versus no hair loss gave 0.690-0.711 with age and slightly lower without. External validation in an early-onset enriched MPB dataset from the Bonn Study (N = 991) showed improved prediction accuracy without considering age such as AUC of 0.830 for no vs. any hair loss. Because of the large number of markers and the large independent datasets used for the different analytical steps, the newly presented genetic prediction models are the most reliable ones currently available for MPB or any other human appearance trait.

Fig. 1. Relationship between the number of SNP predictors and the prediction accuracy expressed as AUC and AIC for 4-category MPB prediction in the marker ascertainment dataset of 55,573 UKBB males.

This dataset includes the 52,874 UKBB men previously applied for discovering the MPB-association of the 270 SNPs [8] used here for feature selection. Based on this dataset, a MLR model was fit in a randomly selected 80% prediction model training subset and applied in the remaining 20% prediction model testing subset with 100 replicates. The estimated mean AUCs are depicted as coloured line and the 5–95% boundary as respective light colour shade for the four predicted MPB categories. The dashed line marks the cut-off at 117 SNPs that were used as best-fit prediction marker set in all subsequent prediction analyses.

Fig. 2. Individual contribution of the ascertained 117 SNP predictors and age on the prediction accuracy expressed as AUC for the 4-category MLR and the 2-category BLR MPB prediction models obtained from the model testing dataset of 26,177 UKBB men.

These prediction models were built in the prediction model training dataset of 104,694 UKKBB men.

Fig. 3. Histograms of predicted probability overlaid with percentage of MPB categories (including slight, moderate and severe hair loss) in each probability bin.

A The probability was derived from MPB prediction model with 117 SNPs and age as predictors in the prediction model testing dataset of 26,177 UKBB men. These prediction results were practically informative for about 26.2% of the tested males (0.4% <0.2 and 25.8% >0.8). B The probability was derived from MPB prediction model with 107 SNPs as predictors without age in the external validation dataset of 991 men from the Bonn Study. These prediction results were practically informative for about 41.8% of the tested males (0.2% <0.2 and 41.6% >0.8). These 117-SNP (A) and the 107-SNP (B) MPB prediction models were built in the prediction model training dataset of 104,694 UKBB men.