Table 2.

Bayesian Machine Learning predictions for chordoma activity for compounds not in the training sets.

Compound	Broad Bayesian score	Broad model applicability	Broad/EGFR Bayesian score	Broad/EGFR model applicability	EGFR Bayesian score	EGFR model applicability
AZD2014	0.46	0.91	0.56	0.96	0.67	0.82
RDEA11	0.62	0.66	0.66	0.67	0.60	0.57
AZD4054	0.53	0.53	0.50	0.60	0.55	0.55

Datasets were named as Broad²¹ and EGFR²⁰ and models are described in Table 1. Model applicability assesses the portion of fragments overlapping with the training set molecules, higher values indicate more fragments overlapping with the training set. The “score” is the prediction score, a measure of probability of activity with higher values being desirable.