I got this reply from Lynn Miller at GCG.
>I was able to recreate this problem with the EPD database.
>On OSF/1 systems I consistantly see core dumps. On other
>UNIX platforms I get varying results.
>>The problem seems to be with the calculation of the Z score.
>The Z score calculation is new starting with FASTA version 2.
>Since the FASTA in version 8.1 of the Wisconsin Package was
>based on Dr. Pearson's version 1.7, and did not include these
>new statistical calculations, it makes sense that the search
>works in this previous version. I apologize that it no longer
>works in the current version. The addition of statistical
>estimates to the fasta output was much-awaited, and has mainly
>had good effects. Unfortunately, on some data sets, these
>calculations can cause problems.
>>I have asked the developer to look more closely to see what is
>causing this specific problem. In the meantime, you will
>not be able to use FASTA in version 9.0 or 9.1 to search
>the EPD database. Instead, I suggest that you use a
>non-statistically-based search program such as WordSearch.
>I personally do not feel that BLAST is a good option for searching
>this small database.
>>You might also add that the statistical calculations in the
>newer version of FASTA were designed for databases containing a
>large number of unrelated sequences, whereas EPD is a small
>database containing many similar sequences.
>>Regards,
>Lynn Miller, Technical Support Coordinator
--
----------------------------------------------------------------------
Aengus Stewart aengus.stewart at icrf.icnet.uk
Bioinformatics Support Tel: +44 (0)171 269 3679
Imperial Cancer Research Fund
Lincoln's Inn Fields, Holborn, London, WC2A 3PX, UK
----------------------------------------------------------------------
