stranding

Determines genome stranding for sequences mapped to a human reference assembly

Motivation

When handling DNA sequences from third-party data vendors it is common to need to determine whether a given sequence lies on the forward or reverse strand of a human genome reference assembly. This is often crucial to determining which allele is the reference allele and which is the alternate allele.

This software package can:

Determine whether a DNA sequence is on the forward or reverse strand of a reference assembly.
Check for strand-flip errors in public datasets such as HapMap
Validate genomic coordinates of SNV nucleotide probes that are mapped to a reference assembly
Determine the reference and alternate alleles of probe sequences on Illumina Infinium assays

All use-cases require a DNA sequence and an educated guess at where that sequence is mapped in either the GRCh37/hg19 or GRCh38/hg38 reference genome assembly.

Algorithm

We use the provided coordinates to fetch the surrounding reference sequence, compute the reverse complement, and score pairwise local alignments between these sequences and the provided query sequence.

If the query sequence aligns to the forward reference sequence then we determine that the query sequence lies on the forward strand of the reference sequence.
If the query sequence aligns to the reverse complement of the reference sequence then we determine that the query sequence lies on the reverse strand of the reference sequence and a strand flip is required to get to forward genome stranding.
If the query sequence aligns to both the forward and reverse complement of the reference sequence then an exception is raised.
If the query sequence aligns to neither the forward nor reverse complement of the reference sequence then an Unstrandable exception is raised.

Installation

pip install stranding

SeqSeek is installed as a dependency to pull the reference sequences but the sequence files themselves are not automatically downloaded by default. Download them with either:

download_build_37

download_build_38

depending on which assembly you require.

Interface

Consider a probe sequence for the famous rs429358 marker in the APOE gene in Illumina TOP stranding:

apoe_probe = 'TGCACCAGGCGGCCGC[A/G]CACGTCCTCCATGTCCGCG'
apoe_5p = 'TACTGCACCAGGCGGCCGC'
apoe_3p = 'CACGTCCTCCATGTCCGCG'

We can visually determine that this probe sequence aligns to the negative strand of the reference by examining the tracks on the dbSNP site.

The interbase (0-based) coordinate for the position being assayed is 45411940 on chromosome 19 of the build 37 reference assembly.

pos = 45411940
chr = 19

We use the SeqSeek constants BUILD37 and BUILD38 to designate the assembly.

from seqseek import BUILD37, BUILD38

We can determine which strand of the reference assembly these flanking sequences lie on as follows:

from stranding import GenomeStranding 
GenomeStranding().strand_flanks(apoe_5p, apoe_3p, BUILD37, chr, pos)
>>> -1

-1 indicates that the sequences align to the reverse/minus strand of the reference assembly.

+1 indicates that the sequences align to the forward/plus strand of the reference assembly.

Alignment tolerance

By default GenomeStranding(...).strand_flanks(...) allows the query sequences to differ slightly from the reference sequence because it is common to observe slight variations in third-party sequence data. This is especially true when working with locus-specific databases that account for specific haplotypes.

GenomeStranding().strand_flanks('AT' + apoe_5p, apoe_3p + 'CG', BUILD37, chr, pos)
>>> -1

The same result is produced with these altered sequences because a suitable alignment is still found at the specified locus.

You can restrict this to find exact alignments only by passing tolerance=1.0 to the constructor:

GenomeStranding(tolerance=1.0).strand_flanks('AT' + apoe_5p, apoe_3p + 'CG', BUILD37, chr, pos)

Alignment offsets

It is also common to find that sequences from third-party data sources don't align exactly where they claim.

Some reasons for this include:

assaying a nearby locus to detect an insertion or deletion
off-by-one errors or discrepancies due to converting to/from interbase and one-based coordinate systems

By default this package requires that the query sequence aligns to the exact location specified within the allowed tolerance.

GenomeStranding().strand_flanks(apoe_5p, apoe_3p, BUILD37, chr, pos - 10, window=10)

Large window sizes will take longer to compute than small windows.

knmkr / stranding