Spider Neurotoxins, Short Linear Cationic Peptides and Venom Protein Classification Improved by an Automated Competition between Exhaustive Profile HMM Classifiers.

Dominique Koua,Lucia Kuhn-Nentwig

doi:10.3390/toxins9080245

Dominique Koua, Lucia Kuhn-Nentwig

Open Access

https://doi.org/10.3390/toxins9080245

Copy DOI

Abstract

Spider venoms are rich cocktails of bioactive peptides, proteins, and enzymes that are being intensively investigated over the years. In order to provide a better comprehension of that richness, we propose a three-level family classification system for spider venom components. This classification is supported by an exhaustive set of 219 new profile hidden Markov models (HMMs) able to attribute a given peptide to its precise peptide type, family, and group. The proposed classification has the advantages of being totally independent from variable spider taxonomic names and can easily evolve. In addition to the new classifiers, we introduce and demonstrate the efficiency of hmmcompete, a new standalone tool that monitors HMM-based family classification and, after post-processing the result, reports the best classifier when multiple models produce significant scores towards given peptide queries. The combined used of hmmcompete and the new spider venom component-specific classifiers demonstrated 96% sensitivity to properly classify all known spider toxins from the UniProtKB database. These tools are timely regarding the important classification needs caused by the increasing number of peptides and proteins generated by transcriptomic projects.

Highlights

Spiders have evolved a very broad range of venom peptides and proteins designed for predatory and defensive purposes [1,2]
In contrast to ArachnoServer, which completely relies on a rational naming system [9] and, sort sequences according to spider taxonomic families, UniProtKB includes a peptide family classification annotation based on InterPro signatures and, to Pfam profile hidden Markov models (HMMs) [10,11]
Sequences were divided into structural conservation groups based (i) on the distribution of cysteine residues; (ii) on the number of amino acid residues between conserved cysteine; and (iii) on amino acid properties

Summary

Introduction

Spiders have evolved a very broad range of venom peptides and proteins designed for predatory and defensive purposes [1,2]. In contrast to ArachnoServer, which completely relies on a rational naming system [9] and, sort sequences according to spider taxonomic families, UniProtKB includes a peptide family classification annotation based on InterPro signatures and, to Pfam profile HMMs [10,11]. This taxonomy-independent family classification complements the naming system and appears useful to characterize new peptides [12]. A closer look at this UniProtKB spider peptide family classification indicates two major problems

Methods

Results

Discussion

Conclusion