Fig. 1

Illustration of the self-training algorithm. Step 1: train a base predictor, \( f \), using \( {\text{G}}_{1} \) and \( {\text{P}}_{1} \) from animals with measured phenotypes. Step 2: predict self-trained phenotypes, \( {\hat{\text{P}}}_{2} \), based on \( {\text{G}}_{2} \) for animals without measured phenotypes. Step 3: combine \( {\text{G}}_{1} \), \( {\text{G}}_{2} \), \( {\text{P}}_{1} \), and \( {\hat{\text{P}}}_{2} \) to train a new predictor, \( f^{*} \). In the testing phase, compare accuracies of \( f \) and \( f^{*} \) on the testing set (\( {\text{R}}_{\text{SL}} \) and \( {\text{R}}_{\text{SSL}} \)). First, predict phenotypes \( {\hat{\text{P}}}_{\text{T}} \) and \( {\hat{\text{P}}}_{\text{T}}^{ *} \) based on \( {\text{G}}_{\text{T}} \) using \( f \) and \( f^{*} \), respectively and second, calculate \( {\text{R}}_{\text{SL}} \) (\( {\text{R}}_{\text{SSL}} \)) as the correlation between \( {\hat{\text{P}}}_{\text{T}} \) (\( {\hat{\text{P}}}_{\text{T}}^{ *} \)) and \( {\text{P}}_{\text{T}} \)