16 January 2021 in Biodiversity and Conservation
25 June 2025 in Nature Ecology & Evolution
Species identification using DNA barcodes has revolutionized biodiversity sciences. However, conventional barcoding methods may lack power and universal applicability across the tree of life. Alternative methods based on whole genome sequencing are hard to scale due to large data requirements. Here we develop a novel DNA-based identification method, varKoding, using exceptionally low-coverage genome skim data to create two-dimensional images representing the genomic signature of a species. Using these representations, we train neural networks for taxonomic identification. Applying a taxonomically verified novel genomic dataset of Malpighiales plant accessions, we optimize training hyperparameters and find the highest performance by combining a transformer architecture with a new modified chaos game representation. Greater than 91% precision is achieved despite minimal input data, exceeding alternative methods tested. We illustrate the broad utility of varKoding across several focal clades of eukaryotes and prokaryotes. We also train a model capable of identifying all species in the Sequence Read Archive of the National Center for Biotechnology Information using less than 10 Mbp sequencing data with 96% precision and 95% recall and robust to sequencing platforms. The varKoding approach offers enhanced computational efficiency and scalability, minimal data inputs robust to sequencing details and modularity for further development in biodiversity science.
27 April 2024 in Molecular Phylogenetics and Evolution
29 May 2025 in Scientific Data
Genome skimming is a promising sequencing strategy for DNA-based taxonomic identification. However, the lack of standardized datasets for benchmarking genome skimming tools presents a challenge in comparing new methods to existing ones. As part of the development of varKoder, a new tool for DNA-based identification, we curated four datasets designed for comparing molecular identification tools using low-coverage genomes. These datasets comprise vast phylogenetic and taxonomic diversity from closely related species to all taxa currently represented on NCBI SRA. One of them consists of novel sequences from taxonomically verified samples in the plant clade Malpighiales, while the other three datasets compile publicly available data. All include raw genome skim sequences to enable comprehensive testing and validation of a variety molecular species identification methods. We also provide the two-dimensional graphical representations of genomic data used in varKoder. These datasets represent a reliable resource for researchers to assess the accuracy, efficiency, and robustness of new tools to varKoder and other methods in a consistent and reproducible manner.
19 September 2022 in Journal of Vegetation Science
16 December 2024 in bioRxiv
Anthropogenic pressures on biodiversity necessitate efficient and highly scalable methods to predict global species distributions. Current species distribution models (SDMs) face limitations with large-scale datasets, complex interspecies interactions, and data quality. Here, we introduce EcoVAE, a framework of autoencoder-based generative models trained separately on nearly 124 million georeferenced occurrences from taxa including plants, butterflies and mammals, to predict their global distributions at both genus and species levels. EcoVAE achieves high precision and speed, captures underlying distribution patterns through unsupervised learning, and reveals interspecies interactions via in silico perturbation analyses. Additionally, it evaluates global sampling efforts and interpolates distributions without relying on environmental variables, offering new applications for biodiversity exploration and monitoring.