About

DeShPet (acronym for Design of Short Peptides) is a web application, based on VPS service by Srce, developed for the design of maximally diversified random peptide libraries using a genetic algorithm. The optimization uses the fitness function based on the monoisotopic mass of amino acids and helps the user to obtain library design options with varying degrees of mass diversity, where the user can choose the design of interest [1]. Originally developed for the one-bead-one-compound peptide libraries, this approach was applied to simplify the synthesis and consequent chromatography coupled to mass spectrometry identification of all the peptides in a single run in the “bottom-up design approach to OBOC libraries” [2]. Moreover, the optimization offers the possibility of designing multiple peptide libraries, where the user is able to choose the amino acids to be included in the library as well as the number of parallel libraries to be designed [1p]. Furthermore, the application will support the design of libraries with specific properties such as hydrophobicity and charge or activities such as antiviral or antimicrobial by using machine learning based prediction as fitness function [2p].

Additionally, the web application offers the possibility to calculate selected peptide features based on the Peptides package written in R that offers property calculations [3]. The properties include isotopic mass, Cruciani polarity, Cruciani hydrophobicity, Cruciani H-bonding, instability index, Boman index, hydrophobicity (scale = 'Eisenberg'), hydrophobic moment (angle = 100, window = 11), aliphatic index, isoelectric point (pKscale= 'Lehninger'), net charge (pH= 7.4, pKscale= 'Lehninger') and the composition (mol%) based on the amino acids properties (tiny, small, aliphatic, aromatic, non-polar, polar, charged, basic, acidic, sulfur and hydroxylic).

Besides physico-chemical properties, DeShPet also performs prediction of antimicrobial activity for any peptide sequence using an antimicrobial prediction model based on sequential properties with a ROC-AUC score of 0.977 [4, 5]. The prediction model was trained and tested on publicly available data that consist of positive instances collected from the DRAMP 2.0 repository [6] and negative instances collected from the UniProt repository [7].

References

[1] D. Kalafatovic; G. Mauša; T. Todorovski; E. Giralt (2019). Algorithm-supported, mass and sequence diversity-oriented random peptide library design, Journal of Cheminformatics, Vol. 11, 11:25, pp. 1 – 15

[2] D. Kalafatovic; G. Mauša; D. Rešetar Maslov; E. Giralt (2020). Bottom-Up Design Approach for OBOC Peptide Libraries, Molecules, Vol. 25 (15), pp. 1 – 15

[3] D. Osorio; P. Rondon-Villarreal; R. Torres (2015). Peptides: A package for data mining of antimicrobial peptides, The R Journal, Vol. 7 (1), pp. 4 – 14

[4] E. Otović; M. Njirjak; D. Kalafatovic; G. Mauša (2022). Sequential properties representation scheme for recurrent neural network-based prediction of therapeutic peptides, Journal of Chemical Information and Modeling, Vol. 62 (12), pp. 2961 - 2972

[5] Github repository for "Sequential properties representation scheme for recurrent neural network based prediction of therapeutic peptides"

[6] X. Kang; F. Dong; C. Shi; S. Liu; J. Sun; J. Chen; H. Li; H. Xu; X. Lao; H. Zheng (2019). DRAMP 2.0, an updated data repository of antimicrobial peptides, Scientific Data, Vol. 6(1), pp. 148.

[7] The UniProt Consortium (2019). UniProt: a worldwide hub of protein knowledge, Nucleic Acids Research, Vol. 47, Issue D1, pp. D506 – D515