How to Detect Point Mutations in NGS Sequences with CodonCode Aligner
This guide walks you through the process of analyzing NGS sequences to identify
point mutations using CodonCode Aligner. Topics covered on this page include
setup requirements, running the analysis, interpreting results,
and working with reference sequences.
Prerequisites
To detect point mutations with CodonCode Aligner, you’ll need a project with aligned
sequences. You can either import externally created alignments, for example SAM files,
or create contigs in CodonCode Aligner by aligning samples to a reference sequence,
or by assembling sequences if no reference sequence is used.
To generate the contig in CodonCode Aligner:
Create a new project.
Add a reference sequence to the project if available. While optional, a
reference sequence with coding sequence annotation allows Aligner to report the biological
effect of mutations, such as amino acid changes. Reference sequences typically are text
sequences in Genbank or EMBL format.
Create the alignment using one of the NGS tools available in CodonCode Aligner,
for example bowtie2.
For smaller projects, you can also use CodonCode Aligner's "Align to Reference" or "Assemble"
functions:
Use "Make Reference Sequence" from the Sample menu to tell
CodonCode Aligner to use the selected text sequence as the reference sequence.
Import the sequences to align into your project.
Build a contig by aligning the imported sequences to the
reference, or by assembling.
Running Mutation Detection
Once you have contigs of your aligned sequences, you can run mutation detection
to identify potential point mutations across the sequences.
CodonCode Aligner will scan the aligned samples at each base in the contig
and generate a summary table of the results.
In the project view, select the contig or contigs you want to analyze.
From the Contig menu, choose Find Mutations.
The analysis typically takes just a few seconds for small to medium-sized projects.
A progress dialog will appear and can be canceled if needed.
When finished, Aligner will open a new window showing a table that lists each position where
mutations were found. For each sample, the table indicates whether the base is homozygous or
heterozygous, and what the effect of the mutation is.
The buttons in the side bar on the left let you control which columns and which tag types are
shown in the table. For NGS SNP detection, the most important tags are the "polymorphism"
tags that Aligner adds to the consensus sequences.
For contigs that have coding sequence annotation, for example from a designated reference sequence,
the predicted amino acid change will be based on the coding sequence annotation of the reference.
If no reference or no coding sequence annotation is present, the prediction
will be based on translation starting at the first base in the contig.
Reviewing Results
In the example above, six SNPs were found: 5 in coding regions, of which 2 are silent mutations and
3 cause an amino acid change; and one outside of coding regions. Let's have a closer look at some
of the columns.
The last two columns in the table show the coverage depth at each position, and the
frequency of the
variant allele. In the example, coverage ranges from 25-fold to 42-fold, and variant allele frequencies
range from 41% to 66%. This indicates that these positions were heterozygous in the sequenced genome;
homozygous variants would show frequency values closer to 100%.
Note that finding homozygous variants
requires (1) that contigs have a reference sequence, and (2) that the reference is used as the
consensus, which can be set in the "Consensus" preferences.
The Probability column gives an estimate of the probability that the observed
variant is due to random sequencing errors. It is based on the summed quality scores for the
observed variant bases, adjusted for total coverage at the position. If you prefer to instead (or
also) the the corresponding log-scaled quality values, you can use the columns button in the side
toolbar to display the quality columns.
To review a specific mutation in detail, double-click any row in the table.
This opens the selected position in the contig viewn (based on your
double-click preferences).
In this example, we see a roughly even number of C and T reads, as we would expect for a
heterozygous SNP. The assembly around the SNP is clean, without many random errors or gaps,
and the depth of coverage is normal. All this indicates that this is a real SNP in the
BRCA1 coding sequence.
If the visual inspection instead shows many gaps or errors, unusual coverage, or
similar problems, this can indicate problematic regions in the alignment. Often,
additional pre-processing like error correction or re-alignment of the initial
contigs can help to get better and more reliable results.
You can exclude problem regions from mutation analysis by adding a "dontGenotype" tag
to the consensus sequence or the reference sequence at this region. To do so, select
the bases in the consensus or reference sequence, go to the "Sample" menu, and choose
"Add Don't Genotype" from the "Tag" submenu. Alternatively, you can also customize the
toolbar and add a button for this function, or use the keyboard shortcut.
To track manual verification of SNPs, you can mark tags as "Confirmed" by selecting
the base in question, and:
right-clicking on the tag, and choosing Confirm tag
from the popup menu
clicking then on the Confirm button in the toolbar (which may require
customizing the toolbar first), or
by using Confirm Tag from the Tag
submenu in the Sample menu, or
by using a keyboard shortcut.
However, such manual verification is more typically done in Sanger re-sequencing projects than
in NGS projects, where data about SNP probability, coverage, and SNP frequency can be used.
How Aligner Analyses NGS Data For Mutations
The "Find Mutations" function in CodonCode Aligner starts by analysing the selected contig(s)
to determine if the contig is from Sanger sequencing or NGS sequencing. To analyze a contig
using NGS analysis, none of the sequences in the contig can have chromatograms, and the
contigs cannot be "contigs of contigs" (contigs aligned directly, for example
by using the "Compare Contigs" tool).
At each position in the contig, Aligner then compares the bases in each read to the reference
sequence (or, for contigs without a reference sequence, the consensus sequence). Any unaligned
regions at the start or end of sequences are ignored. For each base that differs from the
reference base, the probability that the observed difference is due to random error is then
estimated as follows:
The Phred quality scores for all reads that have a given base change are added.
Given the total coverage (n) at this position and the observed number (s) of base changes
(for the given base), the number of possible combinations that would result in the s changes
is calculated (the binomial n over s).
The probability corresponding to the raw quality sum from step 1 is multiplied with the binomial
from step 2 to give the final probability estimate. It can be shown directly as a probability,
or converted to Phred quality values (-10 log10 scaled).
If the quality value from step 3 is below the threshold set in the mutation detection
preferences (default: 50), the change is ignored; if the quality value is at least the
threshold value, then the SNP is recorded.
To give a simple example, let's assume we have a position where 3 out of 10 reads have an A,
while the other reads and the reference have a T. If the 3 reads all have a quality score of 20,
the total quality of the mutated bases is 60. This raw sum would push the position over our
threshold of 50, but we must consider the coverage, too: the higher the coverage, the more likely
it is that we'd see the base change from random errors. For 10-fold coverage,
there are 120 possible combinations of 3 reads. Combining the two numbers drops the quality score
down to 39.2, below our threshold. But if the quality scores for the three mutated reads were
higher, say 25 instead of 20, that would increase the final quality to 54, above our threshold.
Limitations
NGS mutation detection in CodonCode Aligner is only intended for smaller projects, for
example single gene or single exon studies. It it not suited for larger projects, like
whole chromosome or genome studies.
The accuracy of SNP detection is highly dependent on the quality of the input data,
and can give both false positive and false negative results. For best results,
pre-processing of data with tools for error correction, gap re-alignment, and so on
is recommended.
Mutation finding is optimized for detecting homo- and heterozygous SNPs in resequencing projects
using genomic DNA. It is likely to miss low frequency variants (SNPs present only in a small
subset of reads), for example when sequencing mixed samples or low copy transcript variations.
When importing alignments from SAM files, either the SAM files should contain the sequence
of the reference sequence, or a sequence with the exact name of the reference file as listed
in the SAM file must already be present in the project. Having the sequence present in the
project has the advantage that it can contain coding sequence annotation which can be used
to predict amino acid changes from SNPs.
For multi-base and complex mutations, CodonCode Aligner may not combine the mutations into
a single mutation, but instead list the mutations as separate single-base mutations, or
combine the mutations incorrectly,
The "Find Mutations" tool in CodonCode Aligner is for research use only. Results should always
be reviewed by a qualified scientist.
Adding codingSequence and codonStart Tags
If your reference sequence is not annotated, you can manually add tags in the
contig or base view:
To define the coding region, select the relevant bases in the reference sequence,
right-click, and choose "Add Tag...". Then select
codingSequence from the tag type menu.
If the coding sequence does not start at a codon boundary, you can add a
codonStart tag to one of the first three bases. Use the
Notes field to assign a custom base number (e.g.,
basenumber=3301).
If you have a closely related annotated sequence, you can also use
CodonCode Aligner's "Transfer Tag" function to copy existing annotation tags,
based on sequence identity, to the aligned contig or your reference sequence.
Mutation detection will use these tags the next time you run "Find Mutations"
on the contig.
When correctly aligned and annotated, mutation tags will include codon-level effects
where applicable.
