Optimizing Scorpion Toxin Processing through Artificial Intelligence

Abstract

Scorpion toxins are relatively short cyclic peptides (<150 amino acids) that can disrupt the opening/closing mechanisms in cell ion channels. These peptides are widely studied for several reasons including their use in drug discovery. Although improvements in RNAseq have greatly expedited the discovery of new scorpion toxins, their annotation remains challenging, mainly due to their small size. Here, we present a new pipeline to annotate toxins from scorpion transcriptomes using a neural network approach. This pipeline implements basic neural networks to sort amino acid sequences to find those that are likely toxins and thereafter predict the type of toxin represented by the sequence. We anticipate that this pipeline will accelerate the classification of scorpion toxins in forthcoming scorpion genome sequencing projects and potentially serve a useful role in identifying targets for drug development.

Keywords: RNAseq; neural network; python; sodium channel toxins.

Figure 1

Length cut-off effect on training data and tapai performance. (A) Validation accuracy for different peptide truncation/padding lengths. Results of sequence truncation length on the validation accuracy of the toxin model. (B) Confusion matrix showing tapai performance with the complete dataset (validation and testing sets). (C) Confusion matrix showing tapai performance with sequence truncation length to 128 residues. Color intensity in (B,C) represents the percentage of correct classifications for each combination of predicted and actual classes. Four additional toxin models were created with the TV layer truncating or padding to 16, 32, 64, and 256 residues, and trained using the same hyperparameters (Figure S1).

Figure 2

BLAST sequence similarity and tapai performance. (A) The distribution of percentage of similarity and e-values from the initial BLAST analysis plotted as a function of type of toxin (ICK: red, KTx: green, NaTx: blue). (B) Confusion matrix showing the classification performance of tapai in comparison to BLAST similarity predictions.

Figure 3

Multiple sequence alignment (MSA) of transcripts classified by tapai as (A) NaTx with “only insect” affinity, (B) NaTx with “only mammal” affinity (all from buthid scorpions), and (C) calcins (all from iurid scorpions). Top sequences (those with accession numbers) on each MSA were retrieved from the UniProt. Consensus sequence histograms are found below each MSA (red color indicates the conservative cysteine pattern).