The stability of the whole blood basis matrix was verified by comparing results of deconvolution of whole blood using all cell types to results obtained from deconvolution of the same samples with one or two of the cell types omitted. Omission of a few cell types did not grossly alter the results for the other cell types. Reassuringly, those that were altered were the ones that would be expected. For instance, omitting activated dendritic cells from the basis matrix caused resting dendritic cells to be estimated slightly higher in abundance since those two signatures are substantially similar. There are several limitations of microarray deconvolution that bear discussion. One is that it produces answers about the composition of a sample only for a given set of constituents, in this case the immune cell subset profiles that form the basis matrix. This set can be constructed beforehand or determined via bootstrap methods from the mixture data itself, but either way it may be incomplete or inaccurate. We have assembled a fairly comprehensive set of immune cell subsets with which to analyze SLE, but there are likely to be important subsets not Oxysophocarpine profiled here. Two cell types in particular are known to play important roles in SLE but are not included here because they were deemed to be too rare in blood to be reliably detected by this expression analysis approach: regulatory T cells and Benzoylaconine plasmacytoid dendritic cells. Regulatory T cell levels have been found to be higher in SLE patients and to correlate with disease activity, and their interactions with NK cells make them a very interesting and relevant cell type. Plasmacytoid dendritic cells are the major source of type 1 interferons and are thought to be central to this axis of SLE. Although technically intractable here, inclusion of these cells in studies of this kind would likely offer further insights into SLE biology. Memory T cells were also omitted from the SLE analysis even though they were profiled in the proof of concept experiment because we chose to subset T cells for the basis matrix by CD4/CD8 status and activation status. Adding yet another orthogonal factor, memory status, would have multiplied the number of T cell profiles from four to twelve, an extension that would be interesting but is beyond the scope of this work. The conspicuous lack of increase of activated B cells or plasma cells in our survey of SLE samples may be due to dissimilarity between the in vitro stimulated B cells and CD138+ purified plasma cells we used and the populations of plasma cells that produce the autoreactive antibodies that are a hallmark of SLE. We performed both canonical cytokine activation of B cells and B Cell Receptor ligation via anti-IgM antibodies, but neither protocol for B cell activation yielded a signature that was detected at a higher level in SLE than healthy controls. We confirm in the validation samples that the level of the activation marker CD80 is elevated on B cells from some SLE patients and show that this marker is not differentially 
upregulated in our in vitro activated B cell cultures. This resolves the apparent conflict between the results presented here and the general view that B cells are activated in SLE disease, and it underscores the importance of capturing the complexity of relevant cell biology as fully as possible in any set of basis samples to be used for expression deconvolution. Lahdesmaki et al. used a Bayesian approach to deconvolution that avoids a priori definition of basis groups and instead estimates them.