Joining Illumina paired-end reads for classifying phylogenetic marker sequences

Area of Science:

• Microbiology
• Bioinformatics

Background:

• Illumina sequencing of marker genes is common in metagenomics.
• Paired-end (PE) reads often fail to merge, leading to data loss.
• Utilizing only the first read wastes valuable information from the second read.

Purpose of the Study:

• To rigorously evaluate methods for joining PE reads into single reads for taxonomy annotation.
• To assess the impact of read joining on annotation accuracy and data utilization.
• To provide guidance on optimizing PE read usage in metagenomic studies.

Main Methods:

• Simulated data with sequencing errors and real MiSeq PE data were used.
• Two read joining methods were evaluated against top classifiers (RDP, SINTAX, alignment-based).
• Reference sequences were rearranged for alignment-based classifiers; joined reads were compared to original references for word-counting classifiers.

Main Results:

• Read joining generally improved taxonomy prediction accuracy by mitigating sequencing errors.
• Trimming low-quality bases before joining was crucial for optimal results.
• Read joining increased effective data for annotation, aiding in the identification of asthma-associated bacterial genera.

Conclusions:

• Joining PE reads is recommended when merging is limited, enhancing taxonomy annotation.
• Reference sequence rearrangement may be necessary depending on the classifier used.
• Read joining maximizes Illumina PE data utility and relaxes primer constraints for marker gene studies.