This script computes PSSM (possition specific scoring matrix) for a protein sequence given in input file in FASTA format.
Many thanks to Lukáš Jendele, auhor of original project for protein conservation calculation.
- Download BLAST+ suite from here.
- Install/unpack the BLAST+ archive and make sure your directory is PATH.
- Download CD-HIT from here.
- Unpack CD-HIT archive, run make command and make sure cd-hit is in PATH.
- Download SwissProt and UniRef90 databases from here in FASTA format.
- Create a directory where you want to store the databases.
- Add environmental variable BLASTDB={path to the directory}
- cd to this directory and run these commands. This will take quite a lot of time.
zcat {path to uniref90 database .gz file} | makeblastdb -out uniref90 -dbtype prot -title UniRef90 -parse_seqids
zcat {path to swissprot database .gz file} | makeblastdb -out swissprot -dbtype prot -title SwissProt -parse_seqids
- Optional: Now you can delete the .gz database files.
- cd to the project directory
- Run this command:
./FastProtein.sh {path to your fasta file} [-s]
Parameter -s turnes on scaling of all values using sigmoid function.
Every line of output contains 21 values of PSSM matrix. Order or amino acids is ACDEFGHIKLMNPQRSTVWXY.
Values are scaled using sigmoid function if -s parameter was passed.
PSIBLAST is ran with your fasta input on SwissProt database (1 iteration, eval=1e-5).
The result is filtered (min 80% coverage and seq identity between 30% and 95%), then clustered with CD-HIT (default parameters).
If we are left with less then 50 hits, repeat the same for UniRef90 database.
PSSM generated by last iteration of PSI-BLAST is returned in simplier format.