Table 3.

Classification of MS ligands from datasets A-D according to the cross-validation predictions of the model trained including gene expression (MS(woexp):HPA + MS(wexp):INT) and its associated baseline model trained without this feature (MS(woexp + wexp))

		With gene expression (MS(woexp):HPA + MS(wexp):INT)
		Strong binder	Weak binder	Non-binder
Without gene expression (MS(woexp + wexp))	Strong binder	Conserved Binder (CB) (76.16%)	Unimproved Binder (UB) (2.57%)	Lost Binder (LB) (0.02%)
	Weak binder	Improved Binder (IB) (7.77%)	Conserved Binder (CB)	UnClassified peptide (UC) (12.99%)
	Non-binder	Very Improved Binder (VIB) (0.49%)	UnClassified peptide (UC)	UnClassified peptide (UC)

The percentile rank score predictions of these two models were employed to classify the ligands. A peptide is considered to be a strong binder if its %rank score is ≤ 0.5, a weak binder if 0.5 < %rank score ≤ 2 and a non-binder if %rank score >2%. The number between parenthesis indicates the % of ligands in each category, e.g. Conserved Binder = 76.16%, UnClassified peptide = 12.99%.