PSSpred (Protein Secondary Structure prediction)
is a simple neural network training algorithm for accurate
protein secondary structure prediction.
It first collects multiple sequence alignments using
PSI-BLAST. Amino-acid frequence and log-odds data with Henikoff
weights are then used to train secondary structure, separately,
based on the Rumelhart error backpropagation method.
The final secondary structure prediction result is a
combination of 7 neural network predictors from different
profile data and parameters. The program is freely downloadable
at the bottom of this page.
Click PSSpred_v4.tar.bz2 to download
the current version of the PSSpred package (the programs can only
be run in 64 bit Linux operating system).
1. unpack the PSSpred files by "tar -jxvf PSSpred_v4.tar.bz2".
2. PSSpred needs following external files. If you do not have
them installed in your computer, you can download from here:
to download non-redundant sequence database.
3. Run PSSpred by "PSSpred.pl seq.fasta". You need to edit the $db variable
in the script to point to the location of decompressed nr library.
An instruction can be found at the head of the enclosed "PSSpred.pl" file.
Updates of PSSpred program:
PSSpred V4 was released. Updates include:
remove external dependency on PSI-Blast, making $db the only variable that
need to be editted in PSSpred.pl.
PSSpred V3 was released. Updates include
(1). Convert non-standard amino acids to standard ones to avoid program
failure; (2). change script format so that it can adopt new PSI-Blast
version; (3) a bug was fixed for the PSSpred server
(mismatch between old PSSpred and new PSI-Blast format).
This is the last version the dependent on blast 2.6.
PSSpred V2 was released. Updates include
(1). Dimension is extended so that PSSpred can predict secondary structure
for proteins >1000 residues (up to 4000 residues);
(2). A README file is added into the PSSpred package.
Initial release (PSSpred V1).
Renxiang Yan, Dong Xu, Jianyi Yang, Sara Walker, Yang Zhang.
A comparative assessment and analysis of 20 representative sequence alignment methods for protein structure prediction.
Scientific Reports, 3: 2619 (2013).
[PDF] [Support Information]