BLAST-RT-RICO: Protein secondary structure prediction using BLAST and Relaxed Threshold Rule Induction from Coverings

This webpage contains the Perl programs and the protein data used (for training and testing) in the research paper:

L. Lee, J. L. Leopold, and R. L. Frank, “Protein secondary structure prediction using BLAST and Relaxed Threshold Rule Induction from Coverings”.

Data Source:

Most of the web-based data collection activities were conducted around December 2009.
The protein secondary structure data was downloaded from the Protein Data Bank (PDB) website.
Automated queries (using web crawler programs) were conducted using the web-based NCBI/PSI-BLAST search engine.
The CB396 test dataset (CB396 set) was downloaded from the Barton Group website.
The RS126 test dataset (RS126 set) was downloaded from this research paper webpage.

Perl Programs and Data Download (Step by Step):

The programs were written in Perl programming language.
The steps mentioned below are affected by software program design considerations. It is slightly different from the logical steps mentioned in the research paper.

Step 0, Preliminary Data Preparation: Match data from the CB396 set and the RS126 set against latest secondary structure data from the PDB database. <Download Program and Data (60MB .rar file)>

Step 1, Online BLAST, PDB Data Match and Data Preparation: BLAST search for each test protein, match PDB secondary structure data, and prepare data files for rule generation. <Download Program and Data (18MB .rar file)>

Step 2, RT-RICO Rule Generation and prediction: Main rule generation algorithm/program and main prediction algorithm/program. <Download Program and Data (134MB .rar file)>

Step 3, Q3 score calculation. <Download Program and Data (563KB .rar file)>

TABLE - Protein secondary structure prediction using BLAST-RT-RICO approach on RS126 and CB396 test datasets
(Same as Table 6 in the research paper)

	Training Dataset (For An Individual Protein)
For Test Dataset	Max. No. of Proteins	Min. No. of Proteins	Ave. No. of Proteins
RS126	495	1	41.29
CB396	158	1	15.91

For Test Dataset	Max. No. of 5-Residue Segments	Min. No. of 5-Residue Segments	Ave. No 5-Residue Segments
RS126	107,765	35	8,467
CB396	42,938	20	4,480

For Test Dataset	Max. No. of Rules (at 90% threshold)	Min. No. of Rules (at 90% threshold)	Ave. No. of Rules (at 90% threshold)
RS126	89,235	668	21,242
CB396	98,743	379	18,596

	Test Dataset (all protein domains)
Test Dataset	No. of Proteins Using Offline Preprocessing	Total No. of Residues of the Test Dataset	Q3 (%)
RS126	1	23,416	89.93
CB396	9	62,657	87.71