Sequence Alignment, BLAST, GenBank, Mitomap and EMPOP

Sequence alignment begins with two sequences and asks: what arrangement of matches, mismatches, and gaps makes these sequences most similar? Pairwise alignment follows two main algorithmic approaches. Global alignment (the Needleman-Wunsch algorithm, 1970) aligns the full length of both sequences end-to-end, penalising gaps and mismatches but guaranteeing that every position is accounted for. Local alignment (the Smith-Waterman algorithm, 1981) finds the highest-scoring contiguous region of similarity between two sequences, allowing the rest to be ignored. In forensic species identification, where an unknown sequence fragment may correspond to only part of a reference gene, local alignment is nearly always the appropriate choice.

BLAST (Basic Local Alignment Search Tool), developed at NCBI and first published in 1990, is the most widely used implementation of heuristic local alignment. Rather than computing the full Smith-Waterman dynamic programming matrix for every sequence in a database, BLAST first identifies short exact matches (seeds, typically 11 nucleotides for the nucleotide variant blastn), then extends those seeds only if they score above a threshold. This sacrifices completeness against very divergent sequences but reduces search time from polynomial to near-linear, making GenBank-scale searches feasible in seconds. The output includes an E-value (expected number of hits as good as the result by chance alone) and a bit score. A forensic examiner interpreting a BLAST result should report both metrics: the percent identity and E-value, not just the top-hit name.

In the United States, NIST's Organisation of Scientific Area Committees for Forensic Science (OSAC) has not yet published a specific standard for BLAST query interpretation, but the Scientific Working Group for Wildlife Forensics (SWGWILD) guidelines recommend reporting the top five BLAST hits with percent identity and E-value, and noting whether the top hit is unambiguous or whether multiple species share near-identical top scores. In the European Union, ENFSI's Wildlife Working Group guidelines for species identification carry a similar requirement. In India, wildlife forensics conducted under the Wildlife Institute of India's laboratory framework increasingly follows the SWGWILD model; the broader evidential-weight framework connecting sequence identity to court-admissible probability is addressed in random match probability and the likelihood ratio.

Multiple sequence alignment (MSA) arranges three or more sequences simultaneously so that homologous positions in different sequences are stacked in the same column. MSA is the prerequisite for phylogenetic analysis, population-frequency estimation, and haplogroup assignment. The principal MSA tools a forensic bioinformatician uses are Clustal Omega (updated successor to the original ClustalW), MAFFT (Multiple Alignment using Fast Fourier Transform), MUSCLE (Multiple Sequence Comparison by Log-Expectation), and T-Coffee (which improves accuracy by combining pairwise alignments before producing the multiple alignment).

For mitochondrial DNA forensics, MAFFT and Clustal Omega are the most widely cited in peer-reviewed validation studies. MAFFT's FFT-NS-2 iterative mode is fast enough to handle the mtDNA population datasets used for EMPOP queries (tens of thousands of sequences), while its L-INS-i mode provides higher accuracy for smaller, more divergent sets. The ENFSI DNA Working Group's 2019 guidelines on mtDNA analysis list MAFFT explicitly as an acceptable alignment tool for European forensic labs. The US FBI Laboratory's mtDNA guidelines similarly require a documented MSA step before haplogroup assignment and population frequency estimation.

The output of an MSA is a gap-containing alignment file, typically in FASTA or PHYLIP format, where inserted hyphens represent gaps in a particular sequence relative to the consensus. For forensic reporting, the alignment must be assessed for: (1) correct handling of hypervariable regions in the mitochondrial control region (HV1 and HV2), where insertion and deletion polymorphisms create genuine ambiguity about alignment position; (2) confirmation that the reference sequence used (the revised Cambridge Reference Sequence, rCRS, accession NC_012920) is correctly incorporated; and (3) notation of any heteroplasmic positions in the query sequence.

GenBank, maintained by NCBI/NLM in Bethesda, Maryland, is the primary public nucleotide sequence repository and the default BLAST search target. As of 2024, GenBank holds over 10 trillion nucleotide bases across hundreds of millions of records spanning all described kingdoms of life. The database is part of the INSDC, meaning every submission to GenBank is mirrored in real time to ENA (EMBL-EBI, Hinxton, UK) and DDBJ (National Institute of Genetics, Mishima, Japan). A sequence submitted to any one of the three nodes is automatically assigned an accession number visible in all three.

GenBank does not perform independent biological verification of submitted sequences. Submissions pass through automated checks (format validation, vector contamination screening via VecScreen, and basic annotation review) but the scientific accuracy of the taxonomic identification is the submitter's responsibility. This is the single most important QC limitation for forensic use: a sequence deposited with a wrong species name, even in good faith, will appear as a legitimate match in a BLAST search. The forensic response to this limitation is to cross-validate species identifications using BOLD Systems (for COI barcodes), where sequences are submission-validated against reference libraries curated by taxonomy experts.

When submitting a forensic sequence to GenBank (for example, when contributing evidentiary sequence data to a publication or a shared database), submitters must provide: the organism name at the species level, the isolation source, the collection date, the country of collection, and the sequence coordinates corresponding to the annotated gene. NCBI's BioSample framework requires these metadata fields for all genome-scale submissions. For forensic publications requiring GenBank accession numbers as a condition of journal submission, these metadata requirements must be completed before submission review begins, a step that has delayed several forensic wildlife-trafficking publications.

Human mitochondrial DNA forensics requires two distinct computational steps: haplogroup assignment (placing the sample within the phylogenetic tree of human mtDNA diversity) and population frequency estimation (asking how common the observed haplotype is in the relevant reference population). Two databases dominate:

Mitomap (hosted at Emory University, Atlanta, Georgia, US) is the primary repository for pathogenic and polymorphic mtDNA variants. It curates published mtDNA sequences, disease associations, and population polymorphism data, and it includes the complete rCRS reference sequence (accession NC_012920). Mitomap is the authoritative resource for variant interpretation in medical contexts and is increasingly used in forensic identification of degraded human remains, where a clinical mutation database provides context that a population-only database cannot. Mitomap does not, however, provide population-frequency tables formatted for forensic likelihood-ratio calculations.

EMPOP (EMtDNA POPulation database), curated at the Institute of Legal Medicine at Innsbruck Medical University, Austria, is built specifically for forensic mtDNA haplotype frequency estimation. EMPOP currently holds over 50,000 quality-controlled mitochondrial sequences from population studies across more than 130 populations worldwide. Its forensic QC pipeline is the most stringent of any mtDNA resource: submissions pass through the empop quality assurance pipeline (SAM2, a tool that screens for phylogenetic inconsistencies, sequencing artefacts, and nomenclature errors against the PhyloTree human mtDNA phylogeny curated by van Oven and Kayser). Only sequences passing SAM2 are included in the database used for forensic frequency queries.

The ISFG DNA Commission's recommendations on mtDNA analysis, updated in 2014, state that population frequency estimates for forensic court reporting should be derived from EMPOP rather than ad hoc literature searches. The FBI Laboratory uses EMPOP for its US population mtDNA frequency database. The UK Forensic Science Regulator's approved mtDNA methodology references EMPOP as the standard frequency resource. German state forensic laboratories (under BKA coordination) use EMPOP with population-specific subsets for German and European populations.

The Y-Chromosome Haplotype Reference Database (YHRD, coordinated at Humboldt University Berlin and accessible at yhrd.org) is the primary resource for Y-STR haplotype frequency estimation. As of 2024, YHRD holds over 300,000 male haplotypes from more than 1,400 population samples across more than 130 countries, making it the largest curated Y-STR resource in the world. Submissions are accepted through a standardised process requiring: (1) population designation at the country and metapopulation level; (2) haplotype data in the YHRD standard format (a defined panel of Y-STR loci including the minimal haplotype and, for more recent submissions, Yfiler Plus and PowerPlex Y23 panels); (3) documentation of informed consent and institutional ethics approval; (4) confirmation that the sample is unrelated males (YHRD excludes close relatives because they share identical Y-haplotypes by descent, which would artificially inflate haplotype frequencies).

The ISFG DNA Commission's 2014 recommendations on Y-STR analysis specify that Y-STR haplotype frequencies used in court reporting should be derived from YHRD using population samples appropriate to the case. The metapopulation groupings in YHRD (European, North African, Sub-Saharan African, East Asian, South Asian, Latin American, and others) provide the minimum granularity required for a defensible forensic frequency estimate. Country-specific subsets are available for the most common populations.

The quality-control mechanism that distinguishes YHRD from a simple literature aggregation is the "augmented" search, which applies a counting method (the modified ceiling method or the SWGDAM-recommended counting method) to produce a frequency estimate with a 95% confidence interval rather than a point estimate. For low-frequency haplotypes with zero observations in a given population, YHRD applies the count-based minimal estimator (1/N, where N is the number of haplotypes in the database for that population), which is the forensic convention endorsed by the ISFG and accepted in German, Dutch, and UK courts.