Mutual information and variants for protein domain-domain contact prediction

BMC Research Notes

Table 1 Precision for detecting contact versus non-contact residues at 20% recall

Contact prediction
MI variant	Precision contact vs. non-contact
MIc	44.9
MIp	42.3
MIcRA	36.9
MIpRA	35.8
MIp3D	31.8
MIp3DRA	29.4
MIRA	28.4
Random	24.5
MI	24.4
RandomRA	24.4
MI3DRA	23.5
MI3D	19.9

Results are given for the 40 test cases. MI variants are listed in descending order of contact versus non-contact precision, i.e. best to worst classifier of contact residues. The probability of randomly selecting a contact residue from all surface residues is 24.5%. This probability changes to 24.4% when using the reduced alphabet amino acid set because residues are lost as the entropy of their corresponding MSA column reduces to 0.

ISSN: 1756-0500