FastQC

In order to analyse the quality of the raw data generated from next-generation high throughput sequencers, quality control (QC) reports need to be generated. FastQC is a popular tool that provides such a report, with an overview of basic QC metrics and by spotting any problems that originate from the sequencer or the starting library material.

FastQC is available on the Galaxy hub and can be run for the immediate analysis of a small number of FASTQ files using an interactive mode. A non-interactive mode is used for the systematic processing of a large number of files in an analysis pipeline.

Evaluating results

This tool imports data in BAM, SAM, FASTQ or Fastq.gz file formats and runs different QC analyses called modules. The output is a permanent report in an HTML file format that can be viewed in your browser.

The report is presented as graphical or list data, and each result section represents a FastQC module that was run. In addition, there are flag assignments of “Passed” (green tick), “Warn” (orange exclamation mark) or “Fail” (red cross) to the modules, which should be interpreted with caution. The thresholds for flag assignment are based on what type of sequence data was imported, and are specifically tuned for good quality whole-genome shotgun DNA sequencing, while less reliable for other types e.g. mRNA-Seq; targeted amplicon sequencing. Thus, when receiving a “Warn” or “Fail” flag, researchers should consider what the results mean in the context of the particular sample and the type of sequencing that was run, using the evaluation as pointers to where you should concentrate.

Analysis modules of a standard report

• Basic Statistics – this module provides a simple overview of the input file: the file name; file type; encoding; total sequences; filtered sequences; sequence length. This module never flags a warning or failure.
• Per Base Sequence Quality – a BoxWhisker plot is generated that shows an overview of the range of quality values across all bases at each position of the input file. FastQC attempts to automatically determine which encoding method was used and the title of the graph describes this method. Warnings and failures can be raised for this module for specific unmet thresholds. The quality values represented in this graph and those proceeding are known as “Phred Quality Scores”. Phred quality scores, observed on the y-axis plots as shown in the previous module, are used to indicate the measure of base quality in sequencing. Greater values of Phred indicate the high consistency of a sequenced base. A Phred score of 30 indicates the likelihood of finding 1 incorrect base call among 1000 bases. In other words, the precision of the base call is 99.9% and further interpretation can be seen in the table below:

Table 1- The probability values of a Phred quality score and the interpretation of their base-calling accuracy.

• Per Sequence Quality Scores – this module is presented as a line graph and shows whether a subset of your sequences has low-quality values universally. If an overall low quality is detected for a significant proportion of the sequences, this may indicate a systematic problem. This is indicated by warnings and failures being raised for specific unmet thresholds.
• Per Base Sequence Content – this module displays a line graph plotting the proportion of each base position from the input file, for which each of the four normal DNA bases (GATC) has been called. In general, there is little to no difference expected between the bases and they should run parallel to each other. Warnings and failures can be raised for this module when specific thresholds are not met, whereby biases for single bases can indicate contaminating overrepresented sequences or systematic errors during sequencing.
• Per Base GC Content – this module plots a line graph of the C content of each base position from the input file. This line is expected to run horizontally, reflecting the overall GC content of the underlying genome being investigated. However, warnings and failures are raised when specific thresholds are not met and can indicate similar biases stated above.
• Per Sequence GC Content – this module displays the measure of GC content across each sequence from the input file in comparison to a modelled normal distribution. A normal GC content distribution is generally expected with the central peak corresponding to the underlying genome. Unusual shapes or shifted peaks can indicate contamination or systematic bias which may be reflected by warnings and failure flags.
• Per Base N Content – this module functions by substituting a conventional base call with low confidence to an N, and plots the percentage of base calls at each position for which N was called. Warnings and failure flags raised can infer that the analysis was unable to interpret the data to make valid base calls.
• Sequence Length Distribution – in general, high throughput sequencers generate fragments of uniform length, but some may have reads of varying length. This module displays a graph which plots this distribution of fragment sizes, whereby warnings and failures are raised for different or zero lengths.
• Duplicated Sequences – Low levels of duplication can indicate high levels of coverage of targeted sequences, whilst high levels of duplication can indicate enrichment biases. This module counts and plots the relative degree of duplication for every sequence. Warnings and failure flags are raised for this module when non-unique sequences are over-represented according to specific thresholds.
• Overrepresented Sequences – Diverse sets of sequences are generally expected with a single occurrence observed. However, overrepresented sequences can indicate high biological significance, library contamination, or low diversity. This module lists all sequences making up 0.1% of the total, but only from the analysis of the first 200,000 sequences due to computational memory conservation requirements. In addition, the best hits to a database of common contaminants are reported. Warnings and failure flags are raised for specific thresholds of this module.
• Overrepresented K-mers – K-mers refers to all of a sequence’s subsequences of length. This module counts the enrichment of every 5-mer within the sequence library and calculates the expected level this k-mer should be observed. This is based on the entire base content of the library, which is plotted as the observed/expected ratio for the top six hits of enrichment across the reads. Warnings and failure flags for these modules are based on a three-, five- or ten-fold threshold of enrichment at any individual base position.

MultiQC

Due to the nature of FastQC producing reports on a per-sample basis, tools were needed to address the time-consuming and complex process of compiling QC results. MultiQC is a tool that was developed to scan the individual QC reports, creating a single summary report to visualise the combined results across all samples. This tool enables the fast and easy analysis of key statistics as presented in Figure 1 below.

Click here to enlarge the image

Figure 1 – The top of a typical MultiQC report is shown. The general statistics table can be seen with metrics from a number of different tools gathered for each sample that was generated from the FastQC analysis.

MultiQC is also available on the Galaxy hub enabling the aggregation of the FastQC logs generated from your analysis into a single self-contained HTML report, which can be visualised in any modern web browser. With this report, researchers will be able to make accurate comparisons between their samples, minimise the risk of confounding factors through detecting batch effects and, improve the quality of QC and reporting.

For more in-depth reading regarding FastQC and MultiQC, refer to the references.

References

MultiQC: summarize analysis results for multiple tools and samples in a single report

FastQC Manual

FastQC for quality assessment

FastQC Tutorial and FAQ