gordonii or F. nucleatum suggested the reduction in the number of predicted tryptic fragments unique to P. gingivalis would not be sufficient to impact the analysis of more than a small number of proteins. The qualitative peptide level FDR was controlled to approximately 5% for all conditions by selecting PRN1371 purchase a minimum non-redundant spectral count cut-off number appropriate to the complexity of each condition, P. gingivalis alone or the P. gingivalis-F. nucleatum-S. gordonii community. Protein abundance ratio calculations

Protein relative abundances were estimated on the basis of summed intensity or spectral count values [27, 32, 33] for proteins meeting the requirements for qualitative identification described above. Summed intensity refers to the summation of all processed parent ion (peptide) intensity measurements (MS1) for which a confirming CID spectrum (MS2) was acquired according to the DTASelect filter files. For spectral counts, the redundant numbers of peptides uniquely

associated with each ORF were taken from the DTAselect filter table (t = 0). Spectral counting is a frequency measurement that has been demonstrated in the literature to correlate with protein abundance [54]. These two ways of estimating protein relative abundance, that avoid the need for stable isotope labeling, have been discussed in a recent review [27] with specific reference to microbial systems. To calculate protein abundance ratios, a normalization scheme was applied such that the total spectral counts or total intensities for all P. gingivalis proteins in each Selleckchem Stattic condition were set equal for each comparison. This normalization also had the effect of zero centering see more the log2 transformed relative abundance ratios, see Fig. 2 (and also the frequency histograms in Additional file 1: Figs. SF5 and SF6). The normalized data for each abundance ratio comparison was tested for significance using

either a global G-test or a global paired t-test for each condition, the details of which have been published for this type of proteomics data in which all biological replicates are compared against each other [55, 56], and are also described in the explanatory notes [see Additional file 1]. Both of these testing procedures weigh deviation from the null hypothesis of zero abundance change old and random scatter in the data to derive a probability or p-value that the observed change is a random event, i.e. that the null hypothesis of no abundance change is true. Each hypothesis test generated a p-value that in turn was used to generate a q-value as described [24, 32], using the R package QVALUE [26]. The q-value in this context is a measure of quantitative FDR [25] that contains a correction for multiple hypothesis testing. A q cut-off value of 0.01 was used for all ratios reported in Additional file 1: Table ST1. All statistical calculations were done in R (Ver. 2.5.0), using source code that has been published [32, 33, 55].