Alignment-Free Viral Metagenomics: External Validation of K-mer Based Ensemble Classifiers on an Independent Amazonian Dataset
Resumo
The evaluation of alignment-free viral classifiers has been conducted predominantly through internal cross-validation on curated databases, without exposure to novel environmental viromes. This work presents an outof-distribution validation case study of a previously proposed alignment-free pipeline, in which ExtraTrees and XGBoost classifiers operate over k-mer embeddings generated by Word2Vec and fastText. Trained on the ZOVER repository, the models are applied without retraining to a fully independent metagenomic dataset derived from an RNA mosquito virome from the Amazon Basin. The validation set contains 7,624 sequences, of which 82 are homology-confirmed viruses from highly divergent families such as Phasmaviridae and Phenuiviridae, embedded in 7,542 host transcripts under strong class imbalance close to one virus per ninety-two host sequences. Without any retraining, ExtraTrees recovered 84.6% of the viral sequences and XGBoost 82.3%, with specificity around 88% and ROC-AUC near 0.95, classifying every input sequence in seconds on CPU and without gene annotation. In sensitivity, both models are comparable to three established tools, namely DeepVirFinder, VirSorter2 and geNomad, with no detectable difference given the small positive set, although with lower specificity. The results show that classical ensemble classifiers based on k-mer embeddings retain useful discriminative signal on divergent wild viromes, positioning them as a fast, annotation-free option for first-pass metagenomic screening.
Referências
Altschul, S. F., Gish, W., Miller, W., Myers, E. W., and Lipman, D. J. (1990). Basic local alignment search tool. Journal of Molecular Biology, 215(3):403–410.
Asgari, E. and Mofrad, M. R. K. (2015). Continuous distributed representation of biological sequences for deep proteomics and genomics. PLOS ONE, 10(11):e0141287.
Auslander, N. et al. (2020). Seeker: alignment-free identification of bacteriophage genomes by deep learning. Nucleic Acids Research, 48(21):e121.
Bojanowski, P., Grave, E., Joulin, A., and Mikolov, T. (2017). Enriching word vectors with subword information. Transactions of the Association for Computational Linguistics, 5:135–146.
Camargo, A. P. et al. (2024). Identification of mobile genetic elements with geNomad. Nature Biotechnology, 42(8):1303–1312.
Chen, T. and Guestrin, C. (2016). XGBoost: a scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 785–794, New York. ACM.
Fuques, E., Massey, A. L., Qureshi, F., Campos-Silva, J. V., Ferreira da Silva, D. J., Peres, C. A., Levi, T., and Vega Thurber, R. L. (2026). Large-scale metagenomic surveillance study expands the known diversity of RNA viruses in mosquito populations from the Amazon Basin. PeerJ, 14:e20880.
Geurts, P., Ernst, D., and Wehenkel, L. (2006). Extremely randomized trees. Machine Learning, 63(1):3–42.
Guo, J. et al. (2021). VirSorter2: a multi-classifier, expert-guided approach to detect diverse DNA and RNA viruses. Microbiome, 9(1):37.
Kieft, K., Zhou, Z., and Anantharaman, K. (2020). VIBRANT: automated recovery, annotation and curation of microbial viruses, and evaluation of viral community function from genomic sequences. Microbiome, 8(1):90.
Krishnamurthy, S. R. and Wang, D. (2017). Origins and challenges of viral dark matter. Virus Research, 239:136–142.
Mikolov, T., Chen, K., Corrado, G., and Dean, J. (2013). Efficient estimation of word representations in vector space. arXiv:1301.3781.
Nayfach, S., Camargo, A. P., Schulz, F., Eloe-Fadrosh, E., Roux, S., and Kyrpides, N. C. (2021). CheckV assesses the quality and completeness of metagenome-assembled viral genomes. Nature Biotechnology, 39(5):578–585.
Ng, P. (2017). dna2vec: consistent vector representations of variable-length k-mers. arXiv:1701.06279.
Rainio, O., Teuho, J., and Klén, R. (2024). Evaluation metrics and statistical tests for machine learning. Scientific Reports, 14:6086.
Ren, J. et al. (2020). Identifying viruses from metagenomic data using deep learning. Quantitative Biology, 8(1):64–77.
Silveira Júnior, C. A. C. (2026). Identificação de sequências virais em dados meta-genômicos: uma abordagem de alto desempenho utilizando embeddings. Master’s thesis, Universidade Federal de Pelotas, Pelotas. Available at: [link].
Steinegger, M. and Söding, J. (2017). MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets. Nature Biotechnology, 35(11):1026–1028.
Tampuu, A. et al. (2019). ViraMiner: deep learning on raw DNA sequences for identifying viral genomes in human samples. PLOS ONE, 14(9):e0222271.
Wu, L.-Y., Wijesekara, Y., Piedade, G. J., Pappas, N., Brussaard, C. P. D., and Dutilh, B. E. (2024). Benchmarking bioinformatic virus identification tools using real-world metagenomic data across biomes. Genome Biology, 25(1):97.
Zhou, S. et al. (2022). ZOVER: the database of zoonotic and vector-borne viruses. Nucleic Acids Research, 50(D1):D943–D949.
