Benchmark data for identifying multi-functional types of membrane proteins.

Identifying membrane proteins and their multi-functional types is an indispensable yet challenging topic in proteomics and bioinformatics. In this article, we provide data that are used for training and testing Mem-ADSVM (Wan et al., 2016. "Mem-ADSVM: a two-layer multi-label predictor for identifying multi-functional types of membrane proteins" [1]), a two-layer multi-label predictor for predicting multi-functional types of membrane proteins.

Specifications Table

Value of the data

Knowing the functional types of membrane proteins can be helpful to elucidate the biological functions of membrane proteins.

This article presents the first comprehensive dataset that contains non-membrane proteins, single-functional-type membrane proteins and multi-functional-type membrane proteins.

The dataset presented here can be used as an important benchmark dataset to evaluate the performance of membrane-protein predictors.

Data

Using benchmark datasets for evaluating the performance of predictors are of great significance in various domains of bioinformatics [5], [6], [7], [8], [9], [10], such as membrane protein type prediction [11]. However, existing benchmark datasets for predicting membrane proteins are either incomplete or non-stringent. This data article describes a stringent and comprehensive benchmark dataset that comprises non-membrane proteins, single-functional-type membrane proteins and multi-functional-type membrane proteins. All of the benchmark datasets (Dataset II(C) together with Dataset I, Dataset II(A) and Dataset II(B)) are accessible from the link in .

Experimental design, materials and methods

The dataset (we named as ‘Dataset II(C)’) here is a benchmark dataset to evaluate Mem-ADSVM [1], a webserver to identify membrane proteins and their multi-functional types. Dataset II(C) was created based on two previous datasets [5], [8], which we named as Dataset I [5] and Dataset II(A) [8]. First, we retrieved all of the 7965 non-membrane proteins in Dataset I. The procedures to create Dataset I are as follows: (1) select proteins in the UniProtKB/Swiss-Prot database; (2) exclude those protein sequences annotated with “fragment” (3) exclude those protein sequences with less than 50 amino acid residues; (4) remove those protein sequences annotated with ambiguous words, such as “by similarity”, “potential”, “probable”, etc.; (5) remove those sequences which are annotated with “membrane protein” (6) use BLASTCLUST [12] to reduce the sequence similarity to no more than 80%. The procedures for obtaining Dataset II(A) are similar to those for Dataset I except that the former collected membrane proteins instead of excluding them, and the former reduced the sequence identity to 25% instead of 80%. Because the sequence identity of Dataset I (80%) was much higher than that of Dataset II(A) (25%), we used BLASTCLUST to reduce the sequence similarity to 25%, leading to 2009 non-membrane proteins. Then, we combined these 2009 non-membrane proteins with Dataset II(A) (5307 membrane proteins) to constitute Dataset II(C) with a total of 7316 proteins, of which 7126 belong to one type, 185 to two types and 5 to three types. Specifically, the distribution of Dataset II(C) is as follows: (1) 626 single-pass type I, (2) 299 single-pass type II, (3) 42 single-pass type III, (4) 73 single-pass type IV, (5) 2437 multi-pass, (6) 403 Lipid-anchor, (7) 172 GPI-anchor, (8) 1450 peripheral and (9) 2009 non-membrane.

Subject area	Biology
More specific subject area	Bioinformatics/Computational Biology
Type of data	Text
How data was acquired	Process datasets that were obtained by searching against the UniProtKB/Swiss-Prot database with a series of stringent criteria
Data format	Analyzed
Experimental factors	Proteins were manually annotated and were extracted from UniProtKB.
Experimental features	For each protein sequence, its associated gene ontology (GO) information was retrieved by searching a compact GO-term database[2], [3], [4]with its homologous accession number.
Data source location	Hong Kong SAR, China
Data accessibility	The dataset is available with this article andhttp://bioinfo.eie.polyu.edu.hk/MemADSVMServer/datasets.html