Domain-Level Annotation and Conservation of Human Endogenous Retroviruses in the GRCh38 Genome

💡Introduction: This dataset provides a comprehensive, genome-wide annotation of conserved retroviral protein domains within human endogenous retroviruses (HERVs). Using a reproducible pipeline based on RepeatMasker, EMBOSS getorf, HMMER, and InterProScan, we analyzed over 120,000 open reading frames (ORFs) derived from internal HERV regions annotated in the human reference genome (GRCh38). We identified 17,787 retroviral-like domain hits, including domains from reverse transcriptase, RNase H, integrase, protease, Gag, Env, and accessory proteins. Thousands of domains showed moderate to high conservation, with more than 1,000 exhibiting near-complete alignment to reference HMM profiles. We further refined these annotations using InterProScan and Phobius to identify conserved catalytic residues, structural motifs, and transmembrane features—enabling a detailed assessment of the functional potential of each locus. 🔭 Integration into HERVarium This dataset constitutes the internal domain component of HERVarium, a comprehensive reference resource for Human Endogenous Retroviruses (HERVs) that integrates both internal region annotations and LTR regulatory features. HERVarium aims to provide a unified framework for exploring HERV structure, coding potential, and regulatory architecture.The companion dataset containing LTR annotations with U3–R–U5 segments and transcription factor motif maps is released separately and is also part of HERVarium (see https://doi.org/10.5281/zenodo.17602210). 🧬 The dataset includes: BED files with genomic coordinates of annotated domains FASTA files of predicted ORFs and extracted domain sequences InterProScan and Phobius outputs for protein domain and topology annotation Summary tables of domain conservation and co-occurrence patterns Together, these resources offer a high-resolution view of the protein-coding legacy of HERVs and support future research into their roles in host biology, immune modulation, and evolutionary co-option. 📄Description of each file: HERV_internal_sequences_v3.fasta: nucleotide sequences of merged internal HERV regions extracted from GRCh38 using RepeatMasker annotations. HERV_orfs_aa_sequences_v3.fasta: amino acid sequences of predicted ORFs derived from internal HERV regions using EMBOSS getorf. HERV_hmmscan_output_v3.tbl: raw HMMER hmmscan output table showing all detected domain matches against GyDB profiles. HERV_loci_annotated_domains_v3.tsv: filtered table summarizing the best-scoring conserved domain per domain class per ORF, including coverage and e-value. HERV_domains_v3.bed: BED file with genomic coordinates of mapped domains across the genome. HERV_domains_nt_sequences_v3.fasta: nucleotide sequences of domain-mapped regions extracted from the genome. HERV_domains_aa_sequences_v3.fasta: amino acid sequences of domain-mapped regions extracted from the original ORFs. HERV_interproscan_output_v3.zip: full InterProScan output directory with XML, TSV, and additional annotation files for each ORF. HERV_interproscan_cdd_parsed_v3.tsv: parsed summary of Conserved Domains Database (CDD) hits from InterProScan. ENV_phobius_output_v3.txt: output of Phobius transmembrane domain prediction run on ENV domain sequences. HERV_coocurrence_Gag_Pol_Env_v3.xlsx: summary table showing co-occurrence of multiple conserved domains (Gag, Pol, Env) within the same locus. HERVK_conserved_loci_v3.xlsx: list of HERVK loci with high-confidence domain architecture (e.g., full Gag–Pol–Env), including coverage scores. HERVK_conserved_loci_ORFs_annotations_v3.xlsx: detailed annotations for the ORFs in the most conserved HERVK loci. HERV_internal_v3.bed: BED file with genomic coordinates of reconstructed HERV's internal regions. gydb_domains_classification.tsv: A tab-separated table mapping GyDB HMM profile names to their corresponding retroviral gene classes (GAG, POL, ENV, or ACCESSORY). Needed to run the annotation pipeline. combined_gydb.hmm: concatenated and hmmpress-indexed HMM profiles from the GyDB collection for retroviral domain detection. Needed to run the annotation pipeline. GRCh38.primary_assembly.genome.fa.out.tar.xz: RepeatMasker output file used in the original analysis. 💻Code and citation: The dataset was generated using a custom pipeline developed for this study, available at:🔗 https://doi.org/10.5281/zenodo.18326380 If you use this dataset in your research, please cite both the dataset DOI and the corresponding article: Tomàs Montserrat-Ayuso, Aurora Pujol, Anna Esteve-Codina, A comprehensive annotation of conserved protein domains in human endogenous retroviruses, NAR Genomics and Bioinformatics, Volume 8, Issue 1, March 2026, lqag013, https://doi.org/10.1093/nargab/lqag013 Version history v1: Internal HERV fragments were merged using a maximum gap of 100 bp. v2: The merge gap was increased to 150 bp, which improved the recovery of longer internal regions while still avoiding over-merging across distinct loci. All other steps of the pipeline remained unchanged. v3: The merge gap was further increased to 200 bp after noting that bona fide internal regions were still being fragmented at <200 bp. All other steps of the pipeline remained unchanged. v4: Added HERVarium-ready domain files. v5: Removed the HERVarium-ready domain files (now moved to a dedicated repository). Added, for reproducibility, the RepeatMasker output file used in the original analysis (packaged as GRCh38.primary_assembly.genome.fa.out.tar.xz), the combined HMM profile database (combined_gydb.hmm) used by hmmscan, and the TSV file (gydb_domains_classification.tsv) that maps each HMM profile to its corresponding functional class (GAG, POL, ENV, Accessory, or Other). Some subfamilies initially escaped the internal region selection pattern; the code was subsequently corrected, and the current annotation now contains all subfamilies.