Machine learning models accurately predict clades of proteocephalidean tapeworms (Onchoproteocephalidea) based on host and biogeographical data (Supporting Information)

This upload contains supplementary materials for the manuscript entitled "Machine learning models accurately predict clades of proteocephalidean tapeworms (Onchoproteocephalidea) based on host and biogeographical data." Article published in Cladistics (10.1111/cla.12610) Authors: Philippe Vieira Alves (1,6) Reinaldo José da Silva (1) Tomáš Scholz (2) Alain de Chambrier (3) José Luis Luque (4) Anastasiia Duchenko (5,6) Daniel Janies (5,6) Denis Jacob Machado (5,6) Affiliations: (1) São Paulo State University (UNESP), Institute of Biosciences, Section of Parasitology, Botucatu, SP, Brazil. (2) Institute of Parasitology, Biology Centre of the Czech Academy of Sciences, České Budějovice, Czech Republic.(3) Natural History Museum, Department of Invertebrates, Geneva, Switzerland.(4) Federal Rural University of Rio de Janeiro (UFRRJ), Department of Animal Parasitology, Seropédica, RJ, Brazil.(5) University of North Carolina at Charlotte (UNC Charlotte), College of Computing and Informatics, Department of Bioinformatics and Genomics, Charlotte, NC, USA.(6) University of North Carolina at Charlotte (UNC Charlotte), Center for Computational Intelligence to Predict Health and Environmental Risks (CIPHER), Charlotte, NC, USA. Acknowledgments The authors are indebted to the crew of the Laboratório de Parasitologia de Peixes (UFRRJ) and Laboratório de Parasitologia de Animais Silvestres (LAPAS, UNESP) for their invaluable help with collection and dissection of fish. We are also grateful to Emil Hernández -- Universidade Federal do Pará (UFPA), Altamira; Luiz E. Roland Tavares -- Universidade Federal do Mato Grosso do Sul (UFMS), Campo Grande; Ricardo M. Takemoto -- Universidade Estadual de Maringá (UEM), Paraná; Angélica L. Carvalho -- Instituto Federal de Rondônia (IFRO), Ariquemes; Diego R. Sens -- Instituto Federal do Tocantins (IFTO), Araguatins; Miguel A. Mancini -- Universidad Nacional de Río Cuarto (UNRC), Argentina, for providing the facilities during the field trips. We appreciate the helpful discussions with colleagues at the 99th Annual Meeting of the American Society of Parasitologists. This study was financed by the São Paulo Research Foundation (FAPESP; \#2020/05412-9), and the ``Ciência sem Fronteiras'' Brazilian program—visitant researcher modality (No. 135/2012). P.V.A. was supported by FAPESP (fellowships \#2021/12593-2 and \#2023/00714-5). This project would not be possible without the support of UNC Charlotte. The authors are deeply indebted to UNC Charlotte's Department of Bioinformatics and Genomics and UNC Charlotte's CIPHER Research Center for their invaluable assistance. Contents Appendix S1: List of supplementary tables (both in csv and xlsx formats)Appendix S1a. List of newly collected tapeworms and associated metadata. Appendix S1b. Summary data of all tapeworm sequences used in the phylogenetic analysis. Appendix S1c. Codification of characters and associate states for the character optimization analysis. Appendix S1d. States of character for each terminal used in the character optimization analysis. Appendix S1e. Categorial attributes and associated states used in the Random Forest analysis. Appendix S1f. Selected clades used for training the Random Forest algorithm. Appendix S1g. List of synonyms used through the analyses. Appendix S2: Individual character optimization filesAppendix S2a. List of non-ambiguous transformations and their type. Appendix S2b. Detailed character matrix used to generate the TNT file. Appendix S2c. Legend for characters and character states. Appendix S2d. Non-ambiguous synapomorphies of each node. Appendix S2e. Character matrix in TNT file. Appendix S2f. Tree in Newick format containing branch labels (nodes). Appendix S2g. PDF file of the tree including transformations in nodes with three or more terminals. Appendix S3: Alignment and Tree search filesAppendix S3a. Alignment-ready for the tree search in TNT (28S rRNA). Appendix S3b. Alignment-ready for the tree search in TNT (MT-CO1). Appendix S3c. Alignment-ready for the tree search in TNT (TOTAL). Appendix S3d. TNT script for the tree search. Appendix S3e. Most parsimonious trees in Newick format. Appendix S3f. TNT script for summarizing the strict consensus tree. Appendix S3g. TNT script for the branch length assessment. Appendix S3h. TNT script for the branch support and frequency analysis (only arguments for Relative Bremer support, and Jackknife resampling frequency were used). Appendix S3i. Consensus tree in Newick format (no branch lengths or support values). Appendix S3j. Complete consensus tree (including branch lengths, support and frequency values). Appendix S3k. Consensus tree, including Goodman-Bremer and Jackknife values, in PDF format. Appendix S4: Machine Learning filesAppendix S4a. Python3 script for the Random Forest Classifier algorithm, including the perturbation test. Appendix S4b. Raw input dataset. Appendix S4c. Curated input dataset. Appendix S4d. Target clades for prediction. Appendix S4e. Accuracy of the random forest model across ten replication runs, ordered by increasing levels of perturbation. Appendix S4f. List of most important features across 10 replication runs, ordered by increasing levels of perturbation. Appendix S5. Maximum likelihood files and associated data. Appendix S5a. Alignment-ready for the tree search in IQ-TREE2 (28S rRNA). Appendix S5b. Alignment-ready for the tree search in IQ-TREE2 (MT-CO1). Appendix S5c. Partitions for analyzing the gene alignments simultaneously in IQ-TREE2 (concatenated RNA28S + MT-CO1). Appendix S5d. Best evolutionary models, per partition, as chosen by ModelFinder in IQ-TREE2. Appendix S5e. Script for the tree search in IQ-TREE2. Appendix S5f. Script for the branch support assessment in IQ-TREE2. Appendix S5g. Best-scoring maximum likelihood (ML) tree in IQ-TREE2, including branch support. Appendix S5h. Maximum likelihood tree used in the sensitivity test within YBIRÁ. Appendix S5i. Parsimony tree used in the sensitivity test within YBIRÁ. Appendix S5j. Arguments used to run the sensitivity test within YBIRÁ. Appendix S5k. Tree with clades match/mismatch between the maximum likelihood and parsimony reconstructions. Appendix S5l. Maximum likelihood tree with selected clades highlighted by colors. Appendix S5m. List of selected clades sorted by color. Appendix S5n. Input data used to run the machine learning analysis. It includes the target clades. Appendix S5o. Input data used to run the machine learning analysis. It includes the host and biogeographical attributes. Appendix S5p. Output data generated in the machine learning analysis. It includes the model's accuracy per replicate and perturbation level. Appendix S5q. Output data generated in the machine learning analysis. It includes the list of attributes sorted by importance. Appendix S5r. Output data generated in the machine learning analysis. It includes the perturbation thresholds. Appendix S5s. Output data generated in the machine learning analysis. It includes a graph with the impact of clade perturbation on the random forest model performance. Appendix S5t. Comparison between the accuracy levels of the machine learning analyses using both the maximum likelihood- and parsimony-based clades.