Before the FASTQ file
How laboratory decisions shape sequencing results, and why I connect the data with what happened at the bench.
When I review sequencing data, the analysis pipeline is only part of the picture. I also want to understand what went into the library, how it was prepared and what happened during sequencing.
That perspective comes from working across laboratory science and bioinformatics. At Rehman Medical Institute, my clinical genomics work combined whole-exome and targeted-panel sequencing with variant review and clinician-facing reporting. At NIH Islamabad, I work across molecular laboratory workflows, library preparation, sequencing and analysis for genomic surveillance and outbreak investigations.
These experiences connect at a practical point: the data we analyse carry the effects of decisions made before the FASTQ file existed. Three common short-read sequencing examples show why that connection matters.
1. More reads can represent the same starting molecules
A large read count can look reassuring. However, when a library begins with a limited range of original molecules, amplification can generate many copies of the same fragments. Sequencing those copies adds reads without necessarily adding the same amount of independent information.
For me, the useful question is how much distinct evidence the dataset contains for the intended analysis. Repeated observations of one fragment and observations from many independent fragments should not automatically carry the same interpretation.
FastQC's duplication module estimates how frequently sequences are repeated. It does not count original molecules or distinguish PCR duplicates from naturally identical sequences by itself. Its estimates also depend on sampling and sequence matching. Library type matters: repeated sequences can be expected in enriched libraries and in RNA sequencing, where some transcripts are much more abundant than others.
A duplication warning therefore needs the experimental context before it can guide a decision. (FastQC duplication guidance)
2. Fragment size helps explain adapter sequences
The biological fragment inside a sequencing library is often called the insert. When a read extends beyond that insert, sequencing can continue into the adapter on the other side.
The resulting adapter signal in a FASTQ file has a laboratory explanation: the relationship between the fragment-size distribution and the selected read length. An adapter plot becomes easier to interpret when it is considered alongside the library's sizing information.
Adapter trimming can remove the unwanted adapter sequence from a read. It cannot make the original biological fragment longer or recover sequence that was never present in it.
This is the connection I want learners to make when they encounter an adapter warning. Understanding the library helps explain the pattern and supports an appropriate response. (Illumina guidance on short inserts)
3. A productive run can still leave individual samples short of data
Samples sharing a sequencing run do not automatically receive equal numbers of reads. Library quantification and pooling influence how strongly each sample is represented.
An unbalanced pool can leave one sample with abundant data and another with too little for its intended analysis. A high total read count for the run can conceal that difference. Illumina's library-QC guidance illustrates how rebalancing a pool can improve the distribution of reads between samples. (Illumina guidance on library pooling)
Read allocation is also different from coverage across a genome. A sample can receive many reads while particular regions remain poorly represented. This distinction connects directly with my FASTQ to Phylogeny lessons on mapping, depth and breadth of coverage.
I find it useful to keep the questions separate: how much data did each sample receive, and how well do those data support the regions or targets we need to interpret?
Reading QC with laboratory context
Across these examples, a QC pattern is an observation that needs an explanation. Similar patterns can have different causes, and several factors may contribute to the same result.
Before choosing a processing step, I want to connect three pieces of information: the sample and library history, the sequencing-run metrics, and the evidence in the reads and alignments. That is also why discussion between laboratory scientists and bioinformaticians is so useful. Each sees part of the same process.
Connecting the wet lab with the dry lab
My free, ongoing From FASTQ to Phylogeny training series follows this approach. It introduces practical bioinformatics gradually while explaining how files, quality metrics and analysis results relate to sequencing and laboratory work.
The lessons include slides, copyable commands and downloadable practice datasets for wet-lab scientists, students and beginners. You can follow the published lessons in order and work through the exercises at your own pace: