pkaPS: prediction of protein kinase A phosphorylation sites with the simplified kinase-substrate binding model

Biology Direct

Table 6 Summary of the physical terms T_jin the scoring function of the pkaPS predictor.

T _j	Property	Positions	α_ppt,j	Description
T ₁	(+) H, K, R	-3/-2	1.0	Positive charge
T ₂	EISD860102 [29]	-3/-2	0.030	Hydrophilic residues
T ₃	ZIMJ680104 [87]	-6 to -2	0.020	Isoelectric point (positive charge), long range
T ₄	(+) H, K, R; (-) D, E	-6 to -2	0.48	Total charge, long range
T ₅	GEIM800106 [34]	+1	0.070	β-strand preference
T ₆	GEIM800107 [34]	+1/+4	0.040	β-strand preference, compensated
T ₇	HAGECH94_V [88]	+2/+3	0.040	Size restrictions
T ₈	KARP850101 [89]	+3	0.040	Flexibility
T ₉	KARP850101 [89]	-9 to -4	0.040	Minimal linker – flexibility
T ₁₀	EISD840101 [29]	-9 to -4	0.040	Minimal linker – hydrophilicity
T ₁₁	EISD840101 [29]	+4 to +9	0.058	Minimal linker – hydrophilicity
T ₁₂	KARP850101 [89]	+4 to +9	0.058	Minimal linker – flexibility
T ₁₃	CIDH920105 [90]	-18 to -6, +6 to +23	0.040	Avoid buried regions – hydrophilicity
T ₁₄	VINM940101 [30]	-18 to -6, +6 to +23	0.040	Avoid buried regions – flexibility

The table shows the complete list of physical property terms in the score function. The values t_jin equations 12 and 13 are set equal to zero for Gaussian-type terms and equal to 0.1 for fixed penalties T₁ and T₄. The only adjustable parameter per term is the weight α_ppt,j (equation 10). These parameters have been selected so that S_ppt is close to zero for most of the learning set examples. The values α_profile,j(equation 9) for the positions -6...+6 are the following multiples of a normalization factor 0.051: 7, 6, 3, 5, 6, 6, 6, 3, 2, 5, 3, 4, and 1. Initial guesses for the α_ppt,j and α_profile,j parameters have been calculated with linear kernel support vector machines as implemented in the LIBSVM library [91]. These weights have subsequently been rounded to two significant positions and edited manually to avoid non-positive numbers and to achieve close-to-zero T_jvalues for the learning set sequences.

ISSN: 1745-6150