With the multiplexing strategy, sequencing hundreds of samples in one single sequencing run is possible. For the multiplexed sequencing, individual “barcode” sequences, also called index adapters, are added to each DNA fragment during the library preparation. The libraries are then combined into a pool and sequenced on our high-throughput sequencing platforms. After sequencing, each read can be identified and sorted based on its index adapter, according to its sample origin, before the final data analysis. This process is called demultiplexing. This workflow is shown in figure 1. After each read is sorted to its sample origin, the index adapters are trimmed, and the number of raw reads per sample is reported.
During the mapping process, the reads are mapped to a reference sequence. Sequence reads that map nowhere to the reference are counted as unmapped reads. There are multiple reasons why the number of unmapped reads can be high:
- Species divergence: Especially when working with bacterial samples, the bacterial species might show evolutionary divergence from the reference species. Choosing the correct subspecies or strain as the reference genome is vital to avoid high numbers of unmapped reads. Thus, the reference genome should be as closely related to the sequenced species as possible.
- Incomplete reference genome: If the reference genome is incomplete, reads that would map to the missing part of the reference remain unmapped. This could be particularly true for references only available on the “contig” or “scaffold” level. For this reason, we only use references on the “chromosome” or “full genome-representation” level for our analyses.
- Sample contamination: If a sample is contaminated with another organism, the reads do not map to the reference and remain unmapped.
- Bacteriophages: Bacteriophages or prophages can be found in samples. If the genome sequence of the bacteriophage is not included in the reference, the reads cannot be mapped and remain unmapped.
- Plasmids: Like bacteriophages, exogenous plasmids can occur in bacterial samples. These plasmids are not necessarily stated in the reference; thus, plasmid reads remain unmapped.
- Low-quality reads: If the quality of the reads is too low, the mapping algorithm might not identify the reads as mappable to the reference, and they remain unmapped.
- Inappropriate alignment parameters: If the parameters for the algorithm that maps the reads to the reference are chosen wrongly, reads that would generally map to the reference might not map and remain unmapped.
The number of unmapped reads is connected to the number of mapped reads: If the number of unmapped reads is high, the number of mapped reads is low, and vice versa. The mapped reads refer to those reads that align directly to a single region on the reference genome. Among the mapped reads, duplicate and uniquely mapped reads are distinguished.
Duplicate reads are identified and removed during the mapping process. Duplicate reads are multiple sequencing reads that map to precisely the same location on the reference. The duplicate reads originate from the same starting fragment, which was amplified. Thus, duplicate reads offer no additional information and must be removed for the bioinformatic analyses. Failing to remove the duplicate reads can artificially increase the coverage of certain regions, resulting in a potential overrepresentation of SNPs or calling of false variants. The duplication rate is defined as the fraction of duplicate reads over the uniquely mapped reads. High duplication rates might be caused by
- low sample quality,
- reagents from low-input library preparation kits,
- the use of low-complexity libraries in hybrid capture,
- PCR over-amplification during or after the library preparation, or
- over-sequencing.
Consequently, an adequate sample input should be used to minimize the duplication rate, and PCR cycles should be reduced when possible. Additionally, the use of paired-end sequencing can be helpful.
The uniquely mapped reads remain after subtracting the unmapped and the duplicate marked reads. These reads can be assigned to one location within the reference genome. As these reads are unique, they increase the confidence in the coverage of the target region and, thus, in the variant calling.
To summarize, the number of raw reads is the sum of all unmapped and mapped reads, whereby the mapped reads consist of duplicates and uniquely mapped reads.
Figure 1 | Sequencing workflow. With multiplexing, multiple samples can be analyzed in one single sequencing run. For this, the fragments are barcoded with an index adapter and combined in one pool for sequencing. After sequencing, the reads need to be demultiplexed. In this process, they are sorted back according to their sample origin (red and blue) based on their index adapters (yellow and light blue). The number of raw reads is then provided for every sample. Subsequently, the reads are mapped to a reference. Reads that do not map to the reference for multiple reasons are categorized as unmapped reads (dark red). The reads that map to the reference are further distinguished as duplicate marked reads (light red) and uniquely mapped reads (red).