Help

Glossary

The words this documentation uses, in the sense get_MNV uses them. Terms in bold appear as columns or INFO fields in the output; see Output Formats for where each one is written.

The change itself#

• SNV (single-nucleotide variant): One reference base replaced by one other base. The output also calls this a SNP in the Variant Type column.
• MNV (multi-nucleotide variant): Two or more substitutions that fall inside one codon and are read together, so the amino acid comes from the whole codon rather than from each base alone. This is the thing get_MNV exists to find.
• SNP/MNV: A row whose codon holds several substitutions, reported with both readings: the amino acid each substitution would give on its own, and the one the codon gives with all of them. A row typed plain MNV has no separate per-base reading to report.
• Codon: Three consecutive coding bases, read in the transcript's direction, that specify one amino acid. On the minus strand the transcript reads from higher coordinates down, so a codon's first base has the highest coordinate of the three.
• Indel: An insertion or a deletion. A VCF writes both anchored on a base that does not change, so T>TGCT inserts GCT after the anchor and TGCT>T deletes the three bases after it.
• delins: A change that deletes and inserts at once, such as AT>GGG. Neither a pure substitution nor a pure indel.
• Complex indel: An allele whose decomposition holds an indel together with at least one substitution, so it cannot be described as one simple event.
• Anchor: The base a VCF record names to place an indel: the base the inserted sequence follows, or the base before the deleted ones. It is not changed by the indel, which is why get_MNV asks where a record acts rather than where it begins.
• Event class and event components: How get_MNV read the REF/ALT pair: the class is the shape of the whole allele (snp, mnv, insertion, deletion, delins, complex_indel), and the components are the individual changes it decomposes into, written as SNV:110:A>C or INS:200:+GG.

Where the change falls#

• Intergenic: Outside every annotated feature. get_MNV writes the placeholder intergenic in the Gene column for these, which is not the name of a gene, and --exclude-intergenic removes only these rows.
• Frameshift and in-frame: An indel whose length is not a multiple of three shifts the reading frame of everything downstream in that feature; one whose length is a multiple of three leaves it intact and adds or removes whole residues.
• Splice donor, splice acceptor, splice region: The first two bases of an intron, its last two, and the bases immediately around a junction. The first two are the essential sites and score HIGH impact; the surrounding region scores LOW.
• NMD (nonsense-mediated decay): The cell's route for destroying a transcript that terminates early. get_MNV predicts it only where a stop is genuinely premature, using the rule that a stop more than 50 nucleotides before the last exon junction triggers decay.
• Translation table: Which NCBI genetic code to read codons with, chosen by --translation-table. It decides the amino acids, which codons end a protein (TGA stops under table 11, codes tryptophan under table 2 and glycine under table 25) and which can start one: a gene's first codon is methionine when it is one of the table's initiation codons, GTG and TTG included in bacteria.

What the evidence says#

• Haplotype: A set of changes carried by the same molecule. A codon's substitutions form a haplotype only if the same DNA molecule carries all of them.
• Phasing: Placing changes onto molecules. get_MNV never phases on its own: it reads what the caller declared, and counts what the reads show.
• Declared phase: What the input VCF claims, taken from a |-separated GT and its PS phase set. A /-separated genotype is unphased and claims nothing.
• Linkage, D' and r²: How far two changes travel together in the reads, beyond what their individual frequencies would produce by chance. Two common variants meeting often is not evidence of a haplotype; see Linkage.
• Read support and event support: How many reads carry a change. Substitutions are counted per base through a window cache; an indel is counted from the CIGAR of each read, and its counts are written in the Event Reads columns.
• Strand bias: Support that arrives almost entirely from one strand, which usually points at an artefact rather than a variant. Reported as a Fisher exact p-value with --strand-bias-info and filtered with --min-strand-bias-p.

How the consequence is labelled#

• SO term: The Sequence Ontology name for the consequence, such as missense_variant or splice_donor_variant. A row can state more than one at once, joined with an ampersand, when a base is both coding and in a splice region.
• Impact: The severity bucket the SO term falls in: HIGH, MODERATE, LOW or MODIFIER. When a row states two consequences it takes the more severe of the two.
• Grantham distance: A number for how chemically different two amino acids are, reported for a genuine missense so that a conservative change can be told from a drastic one.

Edit this page on GitHub