An AB1 file contains more than a sequence string. It preserves the four fluorescence trace channels, base calls, quality scores, and run metadata produced by a capillary sequencer. Reviewing that evidence is essential before accepting an mtDNA consensus.
The goal of chromatogram QC is not to make every trace look clean. It is to decide which interval is interpretable, which calls are supported, and which questions require resequencing. A numeric quality score helps triage the file, but acceptance remains an evidence decision made in the context of the assay, controls, and study protocol.
Before opening the trace
First verify identity and expected biology. Record the sample identifier, primer name and orientation, amplicon, plate or run, instrument, and original filename. Confirm that the read is expected to cover the claimed mtDNA coordinates. A beautiful trace assigned to the wrong sample or primer is still unusable.
Work from an immutable original AB1. Calculate a checksum at ingest and make reviewed sequences derivatives rather than replacements. If the instrument software exported a sequence or PDF, keep it, but do not treat it as a substitute for the AB1 evidence.
A repeatable review workflow
- Inspect the trace at full length. Look for the usual low-resolution primer-proximal segment, a stable middle interval, and declining separation near the end. Note sudden signal loss, broad peaks, baseline shifts, dye blobs, or a phase change that persists downstream.
- Compare signal and quality together. Phred scores estimate base-call error probabilities: a Q20 call corresponds to an estimated error probability of 1 in 100, while Q30 corresponds to 1 in 1,000 under the caller’s model. They are not calibrated guarantees for every instrument, chemistry, or sequence context.
- Propose usable boundaries. Identify the first and last stretches that meet the laboratory’s quality rule. Then inspect those boundaries visually. A rolling-window rule avoids reacting to one poor base, but it can retain an isolated artifact or remove a short, recoverable interval.
- Review every consequential site. Inspect low-quality calls, differences from the chosen reference, indels, ambiguity codes, and positions that affect a downstream conclusion. View enough neighboring sequence to recognize compression, slippage, or a shifted base-call register.
- Compare independent evidence. Align forward and reverse reads in the same orientation. Check that their analyzed intervals actually overlap and that primer-derived sequence has been removed. Resolve disagreements from the traces, not by choosing the reference allele or expected haplogroup.
- Apply a disposition. Mark the read accepted over a stated interval, accepted with documented ambiguities, or rejected/resequence. Keep the decision separate from the raw call set.
What a questionable position can mean
| Observation | Plausible explanations | Appropriate next check |
|---|---|---|
| One low, broad peak | weak signal, local compression, end-of-read decay | opposite strand, local quality, neighboring peak spacing |
| Two peaks at one position | mixed template, heteroplasmy, contamination, pull-up, noise | controls, strand concordance, replicate extraction or amplification, validated mixture method |
| Clean sequence followed by superimposed peaks | indel mixture, nonspecific product, co-amplification | sequence context, amplicon specificity, independent amplification |
| Repeated extra peaks under strong peaks | spectral pull-up or poor matrix correction | channel pattern, instrument run QC, other samples in the run |
| Abrupt signal collapse | capillary/run failure, depleted product, problematic template | run metrics, repeat injection or sequencing reaction |
Peak-height ratios alone do not establish heteroplasmy. Detection and quantification require a validated assay with defined analytical thresholds, strand and replicate rules, controls, and an evaluation of background at that position. For research outside such a protocol, report the observation conservatively as an unresolved secondary signal.
Trimming without hiding evidence
Define coordinates unambiguously. Software often represents a retained interval as zero-based, half-open [left, right), while reports commonly use one-based inclusive positions. Preserve both the original read index and mapped mtDNA coordinate when possible.
Do not trim simply until the mean quality passes. Review whether:
- the retained interval includes the target position;
- the first retained bases still contain primer sequence or dye artifacts;
- the final retained bases show adequate peak separation;
- an internal low-quality region splits an otherwise good read;
- forward and reverse reads support the same base after trimming.
Trimming is not correction. If an internal call remains unsupported, use the project’s stated ambiguity convention, leave it unresolved, or obtain new data. Every manual substitution should retain the original call, position, evidence consulted, reason, reviewer, and date.
Common failure modes
Trusting a global average. A high mean Phred can coexist with one poor region at the site of interest. Review local evidence.
Calling through a long low-quality tail. Apparent private variants accumulate when peak separation fails. Stop the analyzed interval before the failure rather than editing calls to expectation.
Ignoring strand dependence. A candidate seen on only one strand may be real, but it is not equivalent to concordant independent observations. State the evidence level.
Treating software flags as verdicts. Thresholds are screening rules. Document the rule, then review the chromatogram and controls.
Overwriting the source. A cleaned FASTA cannot reconstruct trace channels, qualities, or the operator’s decisions. Keep source and derivative linked.
Minimum QC record
- Original AB1 identifier, checksum, instrument/run metadata, sample, primer, and orientation
- Base-caller/software identity and version, where available
- Quality rule, rolling-window size, and proposed and accepted trim coordinates
- Reviewed low-quality intervals and all consequential positions
- Forward/reverse agreement and replicate/control evidence
- Ambiguity and manual-edit log, including reviewer and rationale
- Final disposition and sequence derivative identifier
Where GeneFlow fits
On AB1 ingest, GeneFlow uses Biopython to store the called sequence, per-base Phred scores, four analyzed trace channels, instrument string, run date, and JSON-safe ABIF metadata in an AB1 trace record linked to a sample and source file. Its QC task calculates mean, median, percentile, minimum, and maximum Phred summaries. It flags 20-base windows with mean below Q20 and suggests the outer interval whose rolling mean exceeds Q25; the overall label is based on mean Phred.
Those are transparent triage rules, not an assay-specific acceptance decision. GeneFlow does not infer heteroplasmy, reconcile forward and reverse reads, verify sample identity, or apply a laboratory’s validated thresholds. A researcher must inspect trace morphology and controls, decide the accepted interval, and use validated external processing or repeat sequencing when the evidence is insufficient. GeneFlow’s role is to keep the raw trace representation, automated screen, sample, and later outputs reviewable together.
Further reading
- Ewing et al., Base-calling of automated sequencer traces using Phred. I. Accuracy assessment
- Ewing and Green, Base-calling of automated sequencer traces using Phred. II. Error probabilities
- Biopython
SeqIOsupport for ABI trace files - Parson et al., DNA Commission guidelines for mitochondrial DNA typing