The NGS world is full of metrics—both pre-sequencing metrics, such as yield and library size, and post-sequencing metrics, such as coverage, duplicate rates, and mapping rates. And while many of these metrics apply to both DNA sequencing and RNA sequencing, some are unique to RNA-seq, or have different implications in this context. In this two-part article series, we’ll focus on post-sequencing metrics in RNA-seq.

We’ll first discuss important metrics for determining RNA-seq data quality, regardless of the specific downstream analysis: total reads, duplicate reads, % of rRNA reads, mapping rate, and unique genes and transcripts identified. These metrics can help you understand how successful your library prep was, which in turn can save resources and money by showing you where you need to refine your workflow so that you can reduce the percentage of uninformative reads—because sequencing is expensive, and you want every read to matter.

Total Reads

After the sequencing run is complete, the sequencer will provide several high-level metrics. One of these is the total number of reads, or raw reads, which refers to the sum of all reads from the sequencing run. If a pool of multiplexed libraries was sequenced, you’ll also get the number of reads per indexed sample. The total number of reads can show whether the sequencer was underloaded or overloaded, while the number of reads per sample can determine if each sample was evenly represented.

Duplicate reads

The next step is to identify duplicate reads through bioinformatic analysis. In DNA sequencing, duplicate reads are often identified as having the exact same read-start and read-end sequences, suggesting that the identical reads resulted from a PCR duplication. In RNA-seq, however, higher levels of gene expression can also lead to reads that appear to be duplicates but actually represent differential expression; therefore the use of specialized software and analysis pipelines can help differentiate between various types of duplicates and consider factors like gene expression levels. This is especially important where differential gene expression levels—rather than the identification of which genes are expressed  or specific RNA variants—is the goal.

Residual rRNA Reads

In both prokaryotic and eukaryotic cells, the vast majority of RNA content—80 to 98%—is ribosomal RNA, or rRNA. Most RNA-seq studies are focused on gene expression, and thus on messenger RNA (mRNA, the transcripts for protein coding genes), or on specific types of noncoding RNA. Thus, common RNA-seq workflows reduce the amount of rRNA prior to library preparation in order to (1) avoid wasting sequencing reads on rRNA, (2) maximize the number of informative reads, and (3) increase the ability to efficiently detect rare transcripts.

This is typically achieved through the selection of mature mRNAs by capturing them by their polyA tails (which rRNA lacks), or by depleting the rRNA in the input sample through a process called ribodepletion. Following sequencing and preliminary bioinformatic analysis, residual rRNA reads are reported as a percentage of the total reads; depending on the sensitivity required and the method of library preparation, acceptable % residual rRNA can range from 4% to 10%. If the % rRNA reads are higher than expected, then additional troubleshooting may be required to reduce the amount of rRNA before library prep.

 Mapping rate

The mapping rate, or % of reads mapped, refers to the percentage of total reads that align to the reference genome or transcriptome. This is an essential quality control metric, as a high number of unmapped reads can indicate contamination—a key concern in laboratories that process samples from multiple organisms.

Genomic mapping rates can be further subdivided into the portion of reads that map to exons (protein coding regions and other portions of mature mRNAs), introns (mRNA regions typically removed during splicing), or intergenic regions (the source of many noncoding RNAs). Exon mapping rates will be highest for workflows that use mRNA capture, as the reads will be enriched for mature mRNA. In contrast, ribodepleted RNA samples will yield a greater abundance of intronic sequences (from unprocessed, nascent mRNAs), as well as noncoding RNAs, as all RNA species except rRNA remain in the sample prior to library preparation.

Number of genes (and unique transcripts) detected

The number of genes detected refers to the number of unique genomic loci that the mapped RNA reads identify; it indicates how many genes are expressed at detectable levels in the sequenced sample. In other words, it describes the complexity of the RNA library, and presumably of the RNA input sample; this complexity can vary greatly between different organisms, cells, and tissues. For example, blood-derived RNA contains high levels of globin genes, and a bacterial monoculture will likely express far fewer genes than a gut microbiome sample.

The number of unique transcripts identified is a similar metric; however, this metric also takes into account the presence of transcript variants within a sample (such as splice variants or nascent transcripts). This can yield important functional insights about the sample being studied.

These two metrics—genes and unique transcripts identified—can be impacted greatly by how the RNA sample has been handled, and by the library preparation method used; for instance, direct head-to-head comparisons using the same input RNA can yield different results depending upon the library preparation kit used. Thus, it is important to take this into consideration, and—when possible—to use the same library preparation kit and workflow throughout a given study.

In our next article, we’ll address RNA-seq metrics for specific applications, including tumor gene expression profiling.

For Research Use only. Not for use in diagnostic procedures.