Lovain Network of Viral Dark Matter of American Bats

This repository provides the complete computational framework for the identification, structural validation, and taxonomic contextualization of novel RNA virus RdRp sequences. The pipeline is divided into two main stages: Homology and Phylogenetic Validation: Candidate sequences were identified using the palmid workflow against PalmDB v.2023-04-26 via Diamond v.2.0.15 in ultra-sensitive mode. Validated contigs were subjected to multiple sequence alignment (MAFFT v7.520) and maximum-likelihood phylogenetic inference (FastTree v.2.2) under the JTT+CAT model to resolve relationships with known Viral RNA-dependent RNA Polymerases (VDM). Sequence Similarity Network (SSN) & Evolutionary Divergence: To contextualize the viral "dark matter," we implemented a structural comparative approach. Redundancy was collapsed using MMseqs2 (30% identity, 80% coverage), and clusters were converted into Hidden Markov Model (HMM) profiles using HHmake v.3.3.0. A high-sensitivity all-against-all search was conducted to calculate a normalized distance matrix. Following Neri et al. (2022), evolutionary divergence ($D$) was estimated as: $$D = -\ln\left(\frac{S(P_i, s_j)}{S_{max}(P_i)}\right)$$ where $S(P_i, s_j)$ is the observed bit-score and $S_{max}(P_i)$ is the profile's self-score. The resulting weighted graph was analyzed in Python-igraph v.1.0.1, employing the Louvain algorithm for community detection and the Distributed Recursive Graph Layout (DrL) for visualization. Taxonomic novelty was assessed by mapping reference metadata from the Serratus project, allowing for the quantification of established Orders and Families within the derived clusters. The provided SLURM scripts are modularized to facilitate reproducibility on HPC clusters. Users can find the core logic for the distance matrix calculation and network generation within the scripts/02_network_generation and scripts/03_final_figures directories.