Table 3. Feature subsets selected for both classification and regression tasks and the description of respective features.

Task	Selected features	Description
CLASSIFICATION	BPC	Occurrence frequency of basic and positively charged residue (H, K, R)
	Sulfur	Occurrence frequency of sulfur-containing residue (C, M)
	MCBPC	Maximum consecutive basic and positively charged residue (H, K, R)
	logPFR	Protein folding rate in log10 base, predicted using SeqRate
	CL	Occurrence of dipeptide cysteine and leucine
	QD	Occurrence of dipeptide glutamine and aspartic acid
	VE	Occurrence of dipeptide valine and glutamic acid
REGRESSION HIGH	TP	Occurrence of dipeptide threonine and proline
	VT	Occurrence of dipeptide valine and threonine
	T × MCPhe	Occurrence frequency of threonine interacting with maximum consecutive phenylalanine residue
REGRESSION MEDIUM	ER	Occurrence of dipeptide glutamic acid and arginine
	WQ	Occurrence of dipeptide tryptophan and glutamine
	VT	Occurrence of dipeptide valine and threonine
	R × AbsCharge	Occurrence frequency of arginine interacting with absolute charge per residue
	ANC × MCAliphatic	Occurrence frequency of acidic and negatively charged residues interacting with maximum consecutive aliphatic residue (I, L, V, A, G)
	MCCys × pI	Maximum consecutive cysteine residue interacting with isoelectric point (pI)
REGRESSION LOW	F × logPFR	Occurrence frequency of phenylalanine interacting with protein folding rate in log10 base, predicted using SeqRate55
	S × MCNPH	Occurrence frequency of serine interacting with maximum consecutive non-polar and hydrophilic residue (I, L, V, A, G, P)
	y × transmembrane	Occurrence of tyrosine interacting with occurrence of transmembrane, predicted using TMHMM56
	Y × nlogPFR	Occurrence frequency of tyrosine interacting with protein folding rate in natural log base, predicted using SeqRate