SPINGO: a rapid species-classifier for microbial amplicon sequences

Abstract

Background: Taxonomic classification is a corner stone for the characterisation and comparison of microbial communities. Currently, most existing methods are either slow, restricted to specific communities, highly sensitive to taxonomic inconsistencies, or limited to genus level classification. As crucial microbiota information is hinging on high-level resolution it is imperative to increase taxonomic resolution to species level wherever possible.

Results: In response to this need we developed SPINGO, a flexible and stand-alone software dedicated to high-resolution assignment of sequences to species level using partial 16S rRNA gene sequences from any environment. SPINGO compares favourably to other methods in terms of classification accuracy, and is as fast or faster than those that have higher error rates. As a demonstration of its flexibility for other types of target genes we successfully applied SPINGO also on cpn60 amplicon sequences.

Conclusions: SPINGO is an accurate, flexible and fast method for low-level taxonomic assignment. This combination is becoming increasingly important for rapid and accurate processing of amplicon data generated by newer next generation sequencing technologies.

Fig. 1

Comparison of species level classification accuracy for 12 different 16S rRNA gene regions by SPINGO, RDP-Classifier, UCLUST and BLASTn, using 10 fold cross validation. All classifiers were trained on the SPINGO 16S species level database, used k-mer size 8 and 100 bootstraps

Fig. 2

Performance of SPINGO across three different datasets and three amplicon regions (8-mers with 100 sub-samples; confidence estimate cut-offs at the X axis). a Species classification of the SILVA sequences, and b 21 bacterial species from a mock community. Proportion of correctly TPR = TP/(TP + FP), and incorrectly FPR = FP/(TP + FP) assigned sequences. c Stacked relative species abundance and un-stacked proportions of the most abundant Clostridium clusters in a single stool sample. Species from the Clostridiales order as red gradient and Bacteroidales order as blue gradient. Corresponding comparisons for the mother implementation of the RDP-Classifier (d-f), UCLUST (g-i; X axis shows UCLUST similarity cut-offs), and BLASTn (j-l; X axis shows Percent identity). All classifiers were trained on the SPINGO database, using k-mer size 8 and 100 bootstraps

Fig. 3

Species level classification accuracy for SPINGO (k-mer sizes 8,10 and 12, and 100 sub-samples) and RDP-Classifier (only k-mer size 8 and 100 sub-samples due to RAM exhaustion for higher k-mer sizes) as assessed by 10-fold cross validation on a database of 6,690 cpn60 sequences, using the Universal Target region of each sequence for classification

Fig. 4

Comparison of time required to classify 16S rRNA gene V3V5 amplicon reads when trained on the SPINGO database. SPINGO run times using three different k-mer sizes and two different bootstrap values (8-mer with 10 bootraps by default) compared to the other methods all using k-mer 8. Only one CPU was used in all cases