Figure 1.
Design of the BactoChip and its application for microbial species identification and quantification.
(A) Schematic overview of computational design and its output. (A1) Complete genomes from 186 bacterial species were retrieved from the National Centre for Biotechnology Information microbial database. (A2) Gene sequences core to each target species were defined on the basis of sequence conservation within each clade. Red dots represent the distribution of core genes shared by strains within and outside a target clade. (A3) Core genes unique to each target species were selected by sequence alignment against all available archaeal and bacterial sequences. (A4) Oligonucleotide probes were designed for up to 10 identified unique genes for each target bacterial species. Each probe color represents specificity to a defined bacterial species. (B) Experimental design and example data. (B1) DNA from microbial communities was tested on the BactoChip. Green dots represent Cy3 bound to genomic DNA fragments from a sample hybridized to the chip. (B2) Species relative abundances are finally inferred by normalization of the fluorescence signal for each probe and species.
Table 1.
BactoChip specifications for the 19 target genera used in this study.
Figure 2.
Accurate detection of individual bacterial species using the BactoChip.
Bar plot indicates the Areas Under the receiver operating characteristic Curves (AUCs) for detection of 37 individual target species. Heatmap shows intensities of individual probes (red) and postprocessed aggregate species quantification (green). Almost all tested species (94.6%) were identified unequivocally by the specifically designed oligonucleotide probes, both based on raw data (see diagonal) and by comparing against the microbial gold standard (all AUC values >0.96).
Figure 3.
Quantification of the relative abundances of multiple species from the same genus contained within a single sample.
True versus detected relative abundances for each of 6 Staphylococcus species are shown in blue (gold standard) and red (inferred), respectively. Even DNA relative abundances were targeted, with true experimental abundances of cell copy numbers varying due to differences in genome size. Mean Squared Error (MSE) of relative abundance over all predictions was below 0.0019.
Figure 4.
Accurate determination of microbial community composition of up to 15 species at even and staggered abundances.
(A) Bar plots show true (blue) versus predicted (red and green) relative abundances for a 9-species bacterial community with evenly distributed abundances and high or low DNA concentrations. MSE of relative abundance over all predictions was below 0.0006. (B) True (blue) versus predicted (red) abundances for a 15-species evenly distributed community, total MSE <0.0013. (C) Evaluation of the BactoChip’s overall quantitation of relative microbial abundances in all four staggered communities. Each point represents the predicted vs. true relative abundance for one species in one experiment, with total R2>0.75.
Table 2.
Detection of changes in absolute microbial load among samples.
Figure 5.
Application of the BactoChip to the oral microbiome detects native Streptococcus spp. and correctly identified the introduced spike-in species.
A) Species-level averaged intensities linearly correlate with the total measured DNA in the range 0.125–1.000 µg. B) BactoChip detected the presence of Streptococcus equi in the saliva samples of both subjects at abundances higher than 50%. The introduced spike-ins for five different species at different abundances (from about 0.1% to 10% of the total community) were successfully identified in all cases with accurate distinction between different Streptococcus species. Only the species with a relative abundance greater than 1.5% in at least one sample are reported. C) Quantitative evaluation of the predictions for the introduce spike-ins looking in terms of fold change between the abundances within and between samples. All comparison showed strong consistency with the expected fold changes.