Prediction of Protein Functional Domains from Sequences Using Artificial Neural Networks

Table 4.

Comparison of Various Prediction Methods on Selected Domain Groups

Domain type (number in SWISS-PROT 38) Number of errors
Ann (present method) PROSITE PRINTS PFAM
EGF (340) 3 12 108 17
Fibronectin type 3 (241) 2 n.a. 27 7
Trypsin (273) 0 26 0 0
ANK-repeat (119) 0 n.a. n.a. 9
WD-repeat (247) 7 25 15 8
Cytochrome C (78) 8 4 5 3
  • The total number of sequences that contain the given domain type in Swiss-Prot 38. The release numbers of PROSITE, PRINTS, and PFAM are those used by Swiss-Prot 38.

  • Errors for PROSITE, PRINTS and PFAM were determined from the Swiss-Prot annotation (e.g., a sequence annotated as having an EGF repeat but having no cross-reference to a given pattern database was considered not detected by the corresponding method). Thus, this number contains only false negatives. On the other hand, the errors of ANN contain both false positives and false negatives.

  • n.a. = Not available (i.e., the domain type is not included in the corresponding method).

This Article

  1. Genome Res. 11: 1410-1417

Preprint Server