PASS
(Prediction of Autonomously folded units based on Sequence Similarities)

Concept :

PASS rearranges the output file of BLAST searches. PASS lists sequence regions in the query sequence that are conserved among various proteins. According to tests we have performed with structured protein fragments, sequences that are highly conserved among various proteins have a probability above 50% to fold in isolation and have a native structure that can be analyzed(Autonomously Folding Unit).

Parameters:

Most parameters have been fixed, and only two parameters can be changed in the web page version. These are the maximum(default = 1e-7) and minimum BLAST E-value(default = 0). The maximum BLAST E-value determines the farthest sequence and the minimum BLAST E-value determines the closest sequences included in the calculation. When no maximum BLAST E-value is used, remote sequence similarities will be included and it will decrease the reliability of the AFU prediction. The threshold for detection determines the minimum variation in the number of similar sequences required in order to predict a domain boundary at this residue. We observed that a threshold over 5 to 10 sequences shoud be used for a reliable AFU prediction.

Prediction Accuracy:

According to our test calculation, the AFU ends are predicted with an error of +/- 10 to 20 residues. We observed a 45% success rate when an error window of +/- 20 residues.

Notice:

PASS essentially reports changes in the number of similar sequences along the query sequence. Thus, PASS will not work when no similar sequences have been detected. Also, putative AFU predicted using PASS will not be reliable when only a few or remote similarities were detected.

Reference:

Y.Kuroda, K.Tani, Y.Matsuo and S.Yokoyama. 2000. Automated search of natively folded protein fragments for high-throughput structure determination in structural genomics. Protein Science 9:2313-2321