Denis Jacob Machado

During the COVID‐19 pandemic, the need for quick public health action often conflicted with the careful methods required in phylogenetics. To explore this, we reviewed 217 SARS‐CoV‐2 studies published from January 2020 to March 2025 in 121 journals. We found many methodological problems that weaken the reliability and reproducibility of these studies. Key issues include missing outgroup sampling, which affects ingroup topology, tree rooting and how we interpret evolutionary changes and relationships. Another issue is the lack of gene annotations, which can cause characters from one gene to align with those of a different gene. Many studies also misinterpret support or other branch statistics, treating them as proof of clade accuracy instead of as measures of relative evidence. In addition, 91% of the studies do not follow FAIR (Findable, Accessible, Interoperable, and Reusable) data principles, with data and code often unavailable. Finally, we also found a “prestige paradox”: journals with higher impact factors do not necessarily have better methods or transparency. Therefore, we offer simple guidelines for authors, reviewers and editors to improve transparency and FAIR data standards in viral phylogenetics, making writing and reviewing papers easier, ensuring published phylogenetic analyses remain a trustworthy resource for future pandemics.

Recombination is a key driver of RNA virus evolution, yet its extent and evolutionary implications in dengue virus (DENV) remain incompletely understood. We conducted a comprehensive, genome-wide recombination screen across 6,905 complete DENV genomes representing all four serotypes, 82 countries, and eight decades of sampling (1944–2023) retrieved from the Bacterial and Viral Bioinformatics Resource Center. Using seven complementary recombination detection methods implemented in RDP5, we identified 66 recombination events across 53 unique recombinant sequences, of which 29 are newly described. Events included intra-genotypic ( n = 18), inter-genotypic ( n = 32), and inter-serotypic ( n = 16) exchanges spanning 14 genotypes and four continents, with no meaningful serotype-level enrichment (Cramér’s V = 0.054). Recombination was concentrated in non-structural genes, most frequently NS3 (19 events), NS5 (17), and NS2 (12), while the capsid gene contained no recombination events, consistent with strong functional constraint. Single-nucleotide polymorphism analyses confirmed low divergence between recombinants and their inferred parents in both recombinant and non-recombinant regions. Phylogenomic analysis of 6,642 sequences revealed that recombinants cluster significantly closer to their major parents ( p = 8.9 × 10 ⁻⁶ ) and that their removal does not significantly alter tree topology ( p = 0.898), suggesting that the short length of recombinant regions limits phylogenetic conflict. We also introduce RECOSIM, an unsupervised machine-learning tool for recombination detection that achieved higher precision than RDP5 on both simulated (93.4% vs. 80.0%) and empirical (98.1% vs. 39.3%) datasets. Collectively, these results establish recombination as a widespread, pan-serotypic phenomenon in DENV with implications for genomic surveillance, vaccine evaluation, and evolutionary inference.

Advances in High-Performance Computing (HPC) have enabled increasingly complex genomic analyses, including those in phylogenomics. These analyses contribute to understanding the evolution of viruses and pathogens, improving our knowledge of disease transmission, and supporting targeted public health strategies. However, due to the increasing number of tools and processing steps involved, executing these analyses manually, step by step, becomes error-prone and inefficient. To address this challenge, we present HP2NET, a robust framework for reproducible, efficient, and scalable phylogenetic network analysis. HP2NET integrates five workflows based on state-of-the-art tools such as PhyloNetworks and PhyloNet, allowing the analysis of multiple datasets and workflows in a single execution. The framework includes features such as task packaging and data reuse to improve performance and resource utilization in HPC environments. We perform a comprehensive performance evaluation of the software used within HP2NET, identifying bottlenecks and analyzing gains from parallel processing. Data reuse provided up to 15.35% time reduction, for a small dataset, in our experimental environment, while parallel execution of the five pipelines reduced total runtime by up to 90.96% compared to sequential runs. Finally, we validate HP2NET in a real-world case study by analyzing Dengue virus genomes, demonstrating its applicability value for large-scale phylogenetic analyses.

The protein folding problem is the challenge of predicting a protein's three-dimensional structure from its amino acid sequence. This problem has been a central challenge in molecular biology for over fifty years. The advent of AlphaFold, a deep learning system developed by DeepMind, represented a paradigm shift in structural biology by demonstrating near-experimental accuracy in protein structure prediction. This review traces the evolution of the AlphaFold family of models, from its breakthrough performance in CASP14 through the expanded capabilities of AlphaFold 3 and the recent emergence of the proprietary Isomorphic Labs Drug Design Engine (IsoDDE). We examine the architectural innovations that underpin AlphaFold's success, its broad applications in drug discovery, virology, and protein engineering, and its well-documented limitations in modeling intrinsically disordered regions, conformational ensembles, and allosteric mechanisms. We also discuss the growing tension between open science and commercial interests in AI-driven structural biology. The review draws primarily on peer-reviewed literature and curated expert sources to provide an accessible yet rigorous overview of the current state and future trajectory of AI-based protein structure prediction.

The considerable HIV-1 genetic diversity has several implications for viral adaptive and evolutionary capabilities. Its genetic diversity is due to its high mutational rates derived from the error-prone viral reverse transcriptase activity, which generates highly heterogeneous viral populations. Moreover, genetic diversity can also increase due to intra- or intersubtype viral genomic recombination following multiple infections. This study examines HIV-1 intersubtype recombinant viruses and their increased genomic diversity over a 12-month period in five individuals from São Paulo state, Brazil. We collected peripheral blood mononuclear cells once every three months from selected participants at five distinct visits. Molecular clones of 1.15 Kbp fragments of the Pol polyprotein, spanning the protease and a portion of the reverse transcriptase (RT) genes, were generated by bulk PCR. Pol sequences were used for evolutionary analysis, including phylogenetics (using TnT), genetic diversity (using Highlighter), and hypermutation frequency (using Hypermut). Recombination detection experiments were conducted with a jumping profile-hidden Markov model (jpHMM), SimPlot++, and RDP5. We observed great genetic diversity and frequent recombination events in all patients. Furthermore, most of the patients presented hypermutations. These findings highlight the dynamic nature of HIV-1 genetic diversity, driven by frequent recombination and hypermutation, which can accelerate viral adaptation and diversification, underscoring the challenges for treatment, prevention, and disease control.

Dendrobatid poison frogs have become well established as model systems in several fields of biology. Nevertheless, the development of molecular and genetic resources for these frogs has been hindered by their large, highly repetitive genomes, which have proven difficult to assemble. Here we present a draft assembly for Phyllobates terribilis (12.6 Gb), generated using a combination of sequencing platforms and bioinformatic approaches. Similar to other poison frog sequencing efforts, we recovered a highly fragmented assembly, likely due to the genome’s large size and very high repeat content, which we estimated to be ≈88%. Despite the assembly’s low contiguity, we were able to annotate multiple members of three gene sets of interest (voltage-gated sodium channels and Notch and Wnt signaling pathways), demonstrating the usefulness of our assembly to the amphibian research community.

Dendrobatid poison frogs have become well established as model systems in several fields of biology. Nevertheless, the development of molecular and genetic resources for these frogs has been hindered by their large, highly repetitive genomes, which have proven difficult to assemble. Here we present a draft assembly for Phyllobates terribilis (12.6Gb), generated using a combination of sequencing platforms and bioinformatic approaches. Similar to other poison frog sequencing efforts, we recovered a highly fragmented assembly, likely due to the genome’s large size and very high repetitive content, which we estimate to be ≈ 88%. Despite the assembly’s low contiguity, we were able to annotate multiple members of three gene sets of interest (voltage-gated sodium channels and Notch and Wnt signaling pathways), demonstrating the usefulness of our assembly to the amphibian research community.

Proteocephalids are a cosmopolitan and diverse group of tapeworms (Cestoda) that have colonized vertebrate hosts in freshwater and terrestrial environments. Despite the ubiquity of the group, key macroevolutionary processes that have driven the group's evolution have yet to be identified. Here, we review the phylogenetic relationships of proteocephalid tapeworms using publicly available (671) and newly generated (91) nucleotide sequences of the nuclear RNA28S and the mitochondrial MT‐CO1 for 537 terminals. The main tree search was carried out under the parsimony optimality criterion, analysing different gene alignments simultaneously. Interestingly, we were not able to recover monophyly of the Proteocephalidae. Additionally, it was difficult to reconcile the tree with host and biogeographical data using traditional character optimization strategies in two dimensions. Therefore, we investigated if host and biogeographical data can be correlated with the parasite clades in a multidimensional space–thus considering multiple layers of information simultaneously. To that end, we used random forests (a class of machine learning models) to test the predictive potential of combined (not individual) host and biogeographical data in the context of the proteocephalid tree. Our resulting models can correctly place 88.85% (on average) of the terminals into eight representative clades. Moreover, we interactively increased the levels of clade perturbation probability and confirmed the expectation that model accuracy negatively correlates with the degree of clade perturbation. Our results show that host and biogeographical data can accurately predict proteocephalid clades in multidimensional space, even though they are difficult to optimize in the parasite tree. These results agree with the assumption that the evolution of proteocephalids is not independent of host and biogeography, and both may provide external support for our tree.