Figure 4.

Switch-Error Rate for HAPI-UR Phasing on Several Subsets of the CHOP Data Set, Stratified by Ethnic Group and Sample Size

(A) Switch-error rate for 88 African American trio parents either phased combined with other ethnicities (“African Combined”) or alone (“African Alone”) and for 597 European American trio parents phased with other ethnicities (“European Combined”) or alone (“European Alone”). Switch-error rate decreases with sample size and also decreases when we include samples from multiple ethnicities in the phasing run. The full 58,207 data set contains African Americans and European Americans along with Latinos and East Asians, and the switch-error rate is lowest for both African Americans and European Americans with the use of this full data set.

(B) Switch-error rate for 88 African American trio parents in a data set of 5,000 African Americans phased either combined with differing proportions of European Americans (“African Combined”) or alone (“African Alone”). When the number of European American samples is equal to or less than the number of African American samples, the switch-error rate of the African Americans decreases relative to phasing alone. As the number of European American samples grows larger relative to the number of African Americans, the switch-error rate increases. When the ratio of African American to European American samples is 1:2 or 1:4, the switch-error rate remains lower for combined phasing than for phasing alone. For a ratio of 1:7.6 African Americans to European Americans, the switch-error rate of combined phasing is higher than for phasing alone.