Two-Target Quantitative PCR To Predict Library Composition for Shallow Shotgun Sequencing

Abstract

When determining human microbiota composition, shotgun sequencing is a powerful tool that can generate high-resolution taxonomic and functional information at once. However, the technique is limited by missing information about host-to-microbe ratios observed in different body compartments. This limitation makes it difficult to plan shotgun sequencing assays, especially in the context of high sample multiplexing and limited sequencing output and is of particular importance for studies employing the recently described shallow shotgun sequencing technique. In this study, we evaluated the use of a quantitative PCR (qPCR)-based assay to predict host-to-microbe ratio prior to sequencing. Combining a two-target assay involving the bacterial 16S rRNA gene and the human beta-actin gene, we derived a model to predict human-to-microbe ratios from two sample types, including stool samples and oropharyngeal swabs. We then validated it on two independently collected sample types, including rectal swabs and vaginal secretion samples. This assay enabled accurate prediction in the validation set in a range of sample compositions between 4% and 98% nonhuman reads and observed proportions varied between -18.8% and +19.2% from the expected values. We hope that this easy-to-use assay will help researchers to plan their shotgun sequencing experiments in a more efficient way. IMPORTANCE When determining human microbiota composition, shotgun sequencing is a powerful tool that can generate large amounts of data. However, in sample compositions with low or variable microbial density, shallowing sequencing can negatively affect microbial community metrics. Here, we show that variable sequencing depth decreases measured alpha diversity at differing rates based on community composition. We then derived a model that can determine sample composition prior to sequencing using quantitative PCR (qPCR) data and validated the model using a separate sample set. We have included a tool that uses this model to be available for researchers to use when gauging shallow sequencing viability of samples.