The simple method of scanning that has been described in the previous section is relatively fast and allows you the flexibility to generate local alignments for those proteins you think will be interesting from an inspection of the hit list. The advanced scanning option allows the NALL local alignment algorithm to be applied during the scan. This means that if the program finds that a protein shows similiarty to multiple regions of a database protein, then all these regions will be reported. This approach has the drawback that it can generate very large numbers of ``hits'' and these can be uninteresting because they are due to matches between repetetive sequences or hydrophobic runs. NALL scanning is also slower, though only by about a factor of 3.
The length dependent statistics are used to screen out alignments that we would expect to see by chance. Furthermore, a minimum alignment length threshold can be set to improve the scan speed.
If you are interested in short ungapped alignments, then it is best to use the program BLAST from NCBI. BLAST is much faster than SCANPS and is highly tuned for the ungapped alignment problem. If you are interested in alignments much over 50 residues in length, then scanps may offer some advantages. You can get ungapped alignments from scanps by setting the gap penalty very high (e.g. -p100), such alignments should agree reasonably well with BLAST output that used the same pairscore matrix.
In order to process the results of a NALL scan properly, you must have built the indexed databases and be able to run the sortsco program (Section 11).