Search

Supplementary Materials

Peptide-based drugs are promising anticancer candidates due to their biocompatibility, and low toxicity. Particularly, tumor homing peptides (THPs) have the ability to bind specifically to can-cer cells receptors and tumor vasculature. Despite their potential to develop antitumor drugs, there are few available prediction tools to assist the discovery of new THPs. Two webservers based on machine learning models are currently active, the TumorHPD (https://webs.iiitd.edu.in/raghava/tumorhpd) and the THPep (http://codes.bio/thpep), and more recently the SCMTHP (SCMTHP (pmlabstack.pythonanywhere.com), based on scoring card method. Herein, a novel method based on network science and similarity searching implemented in the starPep toolbox (http://mobiosd-hub.com/starpep/) is presented for THPs discovery. The approach leverages from exploring the structural space of THPs with Chemical Space Networks (CSNs) and from applying centrality measures to identify the most relevant and non-redundant THPs sequences within the CSN. Such THPs were considered as queries (Qs) for multi-query similarity searches that applies a group fusion (MAX-SIM rule) model. The resulting multi-query similarity searching models (SSMs) were validated with three benchmarking datasets of THPs/non-THPs. Predictions achieved accuracies ranged from 92.64 to 99.18% and Matthews Correlation Coefficients between 0.894-0.98, outperforming state-of-the-art predictors. The best model was applied to repurpose AMPs from the starPep database as THPs, which were subse-quently optimized for the TH activity. Finally, 54 promising THP leads were discovered, and their sequences were analyzed to encounter novel motifs. These results demonstrate the potential of CSNs and multi-query similarity searching for a rapid and accurate identification of THPs.

Supporting Information for a draft of the article titledExploring the Chemical Space of Antiparasitic Peptides and Discovery of New Promising Leads through a Novel Approach based on Network Science and Similarity Searching. We explain the contents of this material below: Supporting information 1 (SI1):FASTA files of 550, 415, and 405 antiparasitic peptides (APPs) to generate networks for this study. Also, this folder contains FASTA files of the five benchmarking datasets of APPs/non-APPs used to compare the performance of our multi-query similarity searching models (mQSSMs) between them and with the algorithms previously reported in the literature. Supporting information 2 (SI2):MS word file with tables of parameters of similarity threshold analysis for chemical space networks (CSNs) and half-space proximal networks (HSPNs) of APPs; common APPs in the top 50 of most central nodes from CSN and HSPN retrieved by weighted degree, hub-bridge, harmonic, and betweenness centrality measures; the number and community membership percentage of the most central APPs by weighted degree, hub-bridge, harmonic, and betweenness centrality measures; and query sets and the number of queries from the mQSSMs to find novel APPs for CSN, HSPN, and CSN-HSPN. Supporting information 3(SI3):Graphml files of all the networks created in this study. Supporting information 4 (SI4):Excel file with normalized weighted degree, hub-bridge, harmonic, and betweenness centrality measures for CSN and HSPN. Supporting information 5 (SI5):FASTA files of the most central and non-redundant APPs by each type of network and centrality measures, or the query sets. Supporting information 6 (SI6):Excel files with results of mQSSMs (Output Predictions). Supporting information 7 (SI7):This folder has three kinds of files, SI7-A that contains three folders with original results for each mQSSM generated, as well an excel file with statistical parameters. SI7-B is an excel file with the performance parameter of the best 21 mQSSMs proposed, as well as the ranking of these models. Finally, SI7-C and SI7-D are pdf files with results of multiple comparisons of our mQSSMs and with literature algorithms, respectively. Supporting information 8 (SI8): FASTA files of 95 lead compounds and communities obtained from the CSN of these compounds. Supporting information 9 (SI9): PowerPoint file with alignments and sequence logos of 95 lead compounds and communities obtained from the CSN of these peptides.

The selenoenzyme glutathione peroxidase (GPx) like activity of stable organoselenium compounds has been evaluated through the initial rate (ν0) of the reduction reaction of H2O2, Cum-OOH, and t-BuOOH. A Quantitative Structure–Activity Relationships (QSAR) analysis based on different machine learning techniques was performed by employing atom-weighed algebraic maps indexes as descriptors. The predictive capability of the obtained models was statistically validated by mean of the correlation coefficient for adjusting (R2), leave one out cross validation (Q2LOO), and bootstrapping (Q2boot). For the case of H2O2 reduction, a model was obtained with six attributes (M2) and values of R2 = 0.907, Q2LOO = 0.867, and Q2boot = 0.852. For the cum-OOH reduction, a model was obtained with five attributes (M15) with the statistical parameters: R2 = 0.925, Q2LOO = 0.894, and Q2boot = 0.873. For the t-BuOOH reduction, a model with four descriptors (M19) was found with the values of R2 = 0.938, Q2LOO = 0.897, Q2boot = 0.856. The statistical parameters obtained for these three models suggest that they are robust enough with good predictive capability. Finally, screening analysis of some related compounds containing selenium was performed and two possible lead compounds were found (16 and 53), which can be used for the searching of candidates with GPx-like activity.

We define eight new macromolecular indices, and several related descriptors for proteins. The coarse grained methodology used for its deduction ensures its fast execution and becomes a powerful potential tool to explore large databases of protein structures. The indices are intended for stability studies, predicting Φ -values, predicting folding rate constants, protein QSAR/QSPR as well as protein alignment studies. Also, these indices could be used as scoring function in protein-protein docking or 3D protein structure prediction algorithms and any others applications which need a numerical code for proteins and/or residues from 2D or 3D format.