iAMAP-SCM: A Novel Computational Tool for Large-Scale Identification of Antimalarial Peptides Using Estimated Propensity Scores of Dipeptides

Abstract

Antimalarial peptides (AMAPs) varying in length, amino acid composition, charge, conformational structure, hydrophobicity, and amphipathicity reflect their diversity in antimalarial mechanisms. Due to the worldwide major health problem concerning antimicrobial resistance, these peptides possess great therapeutic value owing to their low incidences of drug resistance as compared to conventional antibiotics. Although well-known experimental methods are able to precisely determine the antimalarial activity of peptides, these methods are still time-consuming and costly. Thus, machine learning (ML)-based methods that are capable of identifying AMAPs rapidly by using only sequence information would be beneficial for the high-throughput identification of AMAPs. In this study, we propose the first computational model (termed iAMAP-SCM) for the large-scale identification and characterization of peptides with antimalarial activity by using only sequence information. Specifically, we employed an interpretable scoring card method (SCM) to develop iAMAP-SCM and estimate propensities of 20 amino acids and 400 dipeptides to be AMAPs in a supervised manner. Experimental results showed that iAMAP-SCM could achieve a maximum accuracy and Matthew’s coefficient correlation of 0.957 and 0.834, respectively, on the independent test dataset. In addition, SCM-derived propensities of 20 amino acids and selected physicochemical properties were used to provide an understanding of the functional mechanisms of AMAPs. Finally, a user-friendly online computational platform of iAMAP-SCM is publicly available at . The iAMAP-SCM predictor is anticipated to assist experimental scientists in the high-throughput identification of potential AMAP candidates for the treatment of malaria and other clinical applications.

Progress and challenges for the application of machine learning for neglected tropical diseases

“…Utilizing solely the peptide sequence data, an interpretable scoring card system was employed to pinpoint the antimalarial activity (Charoenkwan et al 2022). The authors trained an iAMAP with SCM-based predictor with eight other conventional supervised classifiers of decision tree (DT), k-nearest neighbor (KNN), logistic regression (LR), multilayer perceptron (MLP), naïve Bayes (NB), partial least squares regression (PLSLR), random forest (RF), and support vector machine (SVM) with nine conventional feature descriptors namely amino acid index (AAindex), PCP, amino acid composition (AAC), composition, transition and distribution (CTD), CTD-composition (CTDC), CTD-distribution (CTDD), CTD-trancomposition (CTDT), dipeptide composition (DPC), and tripeptide composition (TPC).…”

AutoPeptideML: A study on how to build more trustworthy peptide bioactivity predictors

“…Negative Class Partitioning Ref AB [36] Random peptides Homology maximisation [36] ACE [27] Random protein fragments Homology reduction (90%) [27] AC1 [1] Antimicrobial peptides Random [5] AC2 [1] Random protein fragments Random [1] AF [36] Random peptides Homology maximisation [36] AMA1 [10,55] Random peptides Random [10] AMA2 [10,1] Random protein fragments Random [10] AMI [36] Random peptides Homology maximisation [36] AO [31] Experimental + random peptides Homology reduction (90%) [31] AP [57] Random peptides Homology reduction (90% for positives and 60% for negatives) [57] AV [36] Random peptides Homology maximisation [36] BBB [12] Random peptides Homology reduction (90%) [12] DPPIV [6] Random + Bioactive Random [8] MRSA [7] Random peptides Homology reduction (80%) [7] NP [4] Random protein fragments Homology reduction (90%) [11] QS [38] Random peptides Random [52] TOX [53] Random peptides Random [53] TTCA [9] T-cell antigens not associated to disease Random [9] with at least one sequence in the evaluation dataset, which compromises their independence. Looking at the partitioning strategies used to generate the datasets, ACE inhibitor, Antioxidant, Antiparasitic, Anti-MRSA, and Neuropeptides all have been constructed with homology reduction strategies with high threshold 80-90%.…”

“…Antibacterial [18] Random non-antimicrobial peptides from UniProt Approximation to homology maximisation [18] ACE inhibitor [16] Random fragments of non-antihypertensive proteins in SwissProt Homology reduction with CD-HIT at 90% [16] Anticancer 1 [4] Antimicrobial peptides Random [28] Anticancer 2 [4] Random fragments of proteins in SwissProt Random [4] Antifungal [18] Random non-antimicrobial peptides from UniProt Approximation to homology maximisation [18] Antimalarial 1 [12,62] Random non-antiparasitical peptides from UniProt Random [12] Antimalarial 2 [12,4] Random fragments of proteins in SwissProt Random [12] Antimicrobial [18] Random non-antimicrobial peptides from UniProt Approximation to homology maximisation [18] Antioxidant [17] Mix of experimentally validated negatives and random peptides from UniProt with same length distribution as positives Homology reduction with Needleman-Wunch pairwise alignments at 90% [17] Antiparasitic [63] Random non-antiparasitical peptides from Uniprot…”

“…One task that epitomises most of the key issues present in the prediction of biological function from sequence is peptide bioactivity prediction. Most available datasets are small and lacking in explicit negative labels, there is no standard representation method for peptide sequences, nor is there any standardised practice regarding control of sequence homology between training and evaluation sets [4, 3, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18]. Additionally, there is significant demand for custom ML predictors as there are many different bioactivities relevant for diverse applications ranging from drug discovery [19] to understanding metabolic or inflammatory signalling pathways [20, 21].…”

“…The main challenge regarding data acquisition for peptide bioactivity prediction is the lack of explicit negative samples. There is a variety of strategies proposed for addressing this problem that include drawing random fragments from proteins in databases like UniProt or SwissProt [4, 16, 9], drawing actual random peptides from the same databases [18, 26, 12, 4], and using peptides with a specific bioactivity as the negative class [4, 10, 17]. In all cases, there are important tradeoffs regarding the generalizability of the model.…”