Microarray-Assisted mobilome Prospecting (MAmP): a method for determining the discrepancy between the physical genome size and that accounted for by known genes represented on an expanded species-specific microarray. (a) A schematic representation of a microarray-based CGI output for an hypothetical region of the genome in a strain under investigation by the MAmP technique. Scanned raw data are normalized permitting classification of microarray-represented genes as ‘Present (+)’ or ‘Absent (−)’ in individual test strains. (b) The arrows represent the genetic organization of these CDS within E.coli K-12 MG1655, E.coli K-12 W3110, E.coli O157 EDL933, S.flexneri 2a Sf301 or other virulence-associated gene clusters included on the MG1655, W3110 or ShE.coli microarrays. (c) Contiguous CDS classified as ‘Present’ are merged into an IC with intergenic non-coding segments between the contiguous CDS included in the corresponding IC. Each IC was then extended in both directions by lengths equal to half the flanking 5′ and 3′ intergenic segments, as indicated by the double-headed arrows of lengths 0.5x and 0.5y in the examples shown. When the ShE.coli meta-array sequences were used, each probe was mapped onto a single source reference template to allow for the generation of specific ICs. (d) The size of the MVG was calculated as the sum of all IC lengths. Consequently, the size of the non-microarray-borne novel mobilome was estimated as equal to the discrepancy between the pulsed-field gel electrophoresis-determined physical genome size and the MVG size. Figure not drawn to scale.

The distribution of H-values corresponding to 4264 amplicon probes spotted onto the MG1655 microarray obtained following a BLASTN similarity search against the EDL933 chromosomal sequence are shown. Interval-grouped H-values are plotted with the data stratified into the experimental CGI categories of ‘Present’ (3775), ‘Absent’ (400) and ‘Indeterminate’ (89). The selected threshold values for in silico CGI, H₀ = 0.40 and H_x = 0.96, are as indicated. The numbers in the boxes at the top right corner correspond to the ‘Number of CDS’ associated with the two bars that extend beyond the limits of the graph. The inset table shows a direct comparison of experimental CGI data derived by Anjum et al. (18) and in silico CGI data for the CDS classified as ‘Present’ or ‘Absent’ only. The sensitivity (S_n) and specificity (S_p) of identifying genes as ‘Present’ when using the in silico as compared to experimental approach, are shown on the right-hand side.

A flow diagram of the logic steps used for in silico CGI. CDS, P, H and S denote CDS mapped onto specific probes, individual microarray-borne amplicon probes, H-values and Bit scores, respectively. P_i and P_j denote distinct individual probes, with the subscripts i and j identify matching CDS, H-values and Bit scores. The steps involved in the primary similarity search are shown in (a), while the cross-hybridization algorithm that re-categorizes CDS from ‘Putatively Present’ to ‘Present’ or ‘Absent’ are shown in (b). A direct comparison of in silico generated data and true experimental CGI data for E.coli EDL933 obtained using the MG1655 microarray was used to validate the algorithm. The key threshold values, H₀ = 0.40 and H_x = 0.96, were set to optimize the sensitivity and specificity of identifying genes as ‘Present’ when using the in silico as compared to experimental approach (see Figure 3).