X_chr simulation and test

[AprilYUZhang]

No description provided.

[AprilYUZhang]

Here are the results generated by alphapeel on autochromosome and xchr. We expect the performance on xchr to be better than that on autochromosome. only hap_0.5 Marker_accuracies, geno_0.3333333333333333 Marker_accuracies, and seg_prob Marker_accuracies have significant benefit.

multi dosage Marker_accuracies 0.792 0.516 0.944 0.978 0.991 0.983
multi dosage Individual_accuracies 0.911 0.601 0.975 0.991 0.996 0.993
multi geno_0.3333333333333333 Marker_accuracies 0.788 0.516 0.918 0.976 0.987 0.978
multi geno_0.3333333333333333 Individual_accuracies 0.907 0.601 0.962 0.989 0.994 0.99
multi hap_0.5 Marker_accuracies 0.125 0.035 0.489 0.964 0.986 0.976
multi hap_0.5 Individual_accuracies 0.845 0.357 0.914 0.979 0.991 0.984
multi geno_prob Marker_accuracies 0.928 0.686 0.953 0.983 0.993 0.987
multi geno_prob Individual_accuracies 0.902 0.601 0.948 0.981 0.992 0.986
multi phased_geno_prob Marker_accuracies 0.916 0.597 0.956 0.985 0.993 0.988
multi phased_geno_prob Individual_accuracies 0.871 0.454 0.945 0.981 0.992 0.985
multi seg_prob Marker_accuracies 0.743 nan 0.105 0.942 0.957 0.94
multi seg_prob Individual_accuracies 0.698 nan 0.102 0.909 0.915 0.854

multi_x_chr x_chr_dosage Marker_accuracies 0.784 0.577 0.927 0.944 0.951 0.945
multi_x_chr x_chr_dosage Individual_accuracies 0.914 0.608 0.977 0.994 0.998 0.994
multi_x_chr x_chr_geno_0.3333333333333333 Marker_accuracies 0.819 0.58 0.953 0.99 0.995 0.986
multi_x_chr x_chr_geno_0.3333333333333333 Individual_accuracies 0.91 0.6 0.965 0.995 0.997 0.992
multi_x_chr x_chr_hap_0.5 Marker_accuracies 0.831 0.458 0.971 1.0 1.0 1.0
multi_x_chr x_chr_hap_0.5 Individual_accuracies 0.858 0.388 0.923 0.993 0.996 0.989
multi_x_chr x_chr_geno_prob Marker_accuracies 0.939 0.705 0.978 0.99 0.996 0.989
multi_x_chr x_chr_geno_prob Individual_accuracies 0.91 0.604 0.972 0.99 0.996 0.99
multi_x_chr x_chr_phased_geno_prob Marker_accuracies 0.928 0.632 0.978 0.991 0.997 0.99
multi_x_chr x_chr_phased_geno_prob Individual_accuracies 0.897 0.537 0.97 0.991 0.996 0.991
multi_x_chr x_chr_seg_prob Marker_accuracies 0.837 nan 0.431 0.959 0.98 0.966
multi_x_chr x_chr_seg_prob Individual_accuracies 0.71 nan 0.056 0.91 0.929 0.899

[AprilYUZhang]

@gregorgorjanc, here is the simulation error example I mentioned. This didn't have anything wrong with it in principle, just means there's room to improve. It follows all of the rules, but has a genotype error. The core issue refers to the peel forward and backwards; during this process, it ignores the information that the "known genotype" provides. Instead, it gets the Segregation probability with error tendency. It is hard to explore what caused it, but imply that not enough attention was paid to known genotypes in the process of peeling.

[XingerTang]

@gregorgorjanc While assessing the accuracy, @AprilYUZhang used different datasets generated by two separate R scripts, which makes direct one-to-one comparisons not necessarily fair. In general, the results are as expected, that introducing x_chr would improve the accuracy.

I plotted the original multi-locus peeling results as red/orange lines, and the multi-locus with x_chr results as blue/purple lines. We can see that there are no significant differences, but the blue lines are higher in general.

If we separate by output files,