Distribution of P-values from QTL mapping of brain RNA expression. The P-value for the best QTL for each microarray probe set is plotted against the rank of that P-value among all probe sets . The four figures show four methods of pre-processing the data before QTL mapping. In each panel the smaller inset shows the entire range of P-values and the larger figure shows the smallest 200 to 300 P-values. The dashed line in the inset shows the expected distribution for random P-values; in the larger figure the dashed line shows the limit for 20% false-discovery rate, according to Benjamini and Hochberg . (a) HWT1 weighting of PM values only; (b) HWT1 weighting of PM-MM differences; (c) HWT1 weighting of PM and MM values combined; (d) PM-MM differences averaged by Affymetrix MAS 5.0 software.