Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2013 Jul 10:14:221.
doi: 10.1186/1471-2105-14-221.

Alternate virtual populations elucidate the type I interferon signature predictive of the response to rituximab in rheumatoid arthritis

Brian J Schmidt  1 , Fergal P Casey, Thomas Paterson, Jason R Chan
Affiliations

Alternate virtual populations elucidate the type I interferon signature predictive of the response to rituximab in rheumatoid arthritis

Brian J Schmidt et al. BMC Bioinformatics. .

Abstract

Background: Mechanistic biosimulation can be used in drug development to form testable hypotheses, develop predictions of efficacy before clinical trial results are available, and elucidate clinical response to therapy. However, there is a lack of tools to simultaneously (1) calibrate the prevalence of mechanistically distinct, large sets of virtual patients so their simulated responses statistically match phenotypic variability reported in published clinical trial outcomes, and (2) explore alternate hypotheses of those prevalence weightings to reflect underlying uncertainty in population biology. Here, we report the development of an algorithm, MAPEL (Mechanistic Axes Population Ensemble Linkage), which utilizes a mechanistically-based weighting method to match clinical trial statistics. MAPEL is the first algorithm for developing weighted virtual populations based on biosimulation results that enables the rapid development of an ensemble of alternate virtual population hypotheses, each validated by a composite goodness-of-fit criterion.

Results: Virtual patient cohort mechanistic biosimulation results were successfully calibrated with an acceptable composite goodness-of-fit to clinical populations across multiple therapeutic interventions. The resulting virtual populations were employed to investigate the mechanistic underpinnings of variations in the response to rituximab. A comparison between virtual populations with a strong or weak American College of Rheumatology (ACR) score in response to rituximab suggested that interferon β (IFNβ) was an important mechanistic contributor to the disease state, a signature that has previously been identified though the underlying mechanisms remain unclear. Sensitivity analysis elucidated key anti-inflammatory properties of IFNβ that modulated the pathophysiologic state, consistent with the observed prognostic correlation of baseline type I interferon measurements with clinical response. Specifically, the effects of IFNβ on proliferation of fibroblast-like synoviocytes and interleukin-10 synthesis in macrophages each partially counteract reductions in synovial inflammation imparted by rituximab. A multianalyte biomarker panel predictive for virtual population therapeutic responses suggested population dependencies on B cell-dependent mediators as well as additional markers implicating fibroblast-like synoviocytes.

Conclusions: The results illustrate how the MAPEL algorithm can leverage knowledge of cellular and molecular function through biosimulation to propose clear mechanistic hypotheses for differences in clinical populations. Furthermore, MAPEL facilitates the development of multianalyte biomarkers prognostic of patient responses in silico.

PubMed Disclaimer

Figures

Figure 1
Figure 1
Scope of the biology represented in the biosimulation platform. In the synovial tissue, the activity of fibroblast-like synoviocytes (FLS), B cells (B), plasma cells, natural killer cells (NK), macrophages (MΦ), and endothelial cells (EC) are modelled. T cells (T) are modelled with distinct CD4 and CD8 T cell subpopulations. CD4 T cells are further divided into Th1, Th2, Th17, and Treg subpopulations, as well as a CD28- subset that has lost the requirement for costimulation. Antigen presentation by dendritic cells (DCs) is also represented. Bone remodelling by osteoclasts (OC) and osteoblasts (OB) as well as cartilage remodelling by chondrocytes and MMPs is included.
Figure 2
Figure 2
Relation of mechanistic axes to measured population-level therapeutic responses. The MAPEL algorithm for developing VPops links the mechanistic axes underlying biosimulation to clinical statistics through VPs. In the bottom panel, MAPEL assigns probability distributions directly to the mechanistic axes. A simple case of two axes is shown for clarity. In the middle panel, prevalence weights for VPs in the cohort are calculated. The prevalence weight is essentially a measure of the fraction of the total VPop that a given VP statistically represents. The prevalence weight of each VP is calculated from the axes weights assigned by MAPEL. VPs assigned a higher weight are depicted as darker colors. In the top panel, the VPop’s clinical response distribution is calculated. As described in the text, the Entelos RA Physiolab® platform was re-run for each individual VP in the population for each therapy. Therefore, biosimulation results provide the response to therapies for each VP. These simulated responses to therapy were used in combination with the prevalence weights to calculate population-level responses to therapy. The calculation of a binned response distribution is shown, e.g. ACR20, 50, 70. In addition, weighted means and weighted standard deviations are also calculated, as detailed in the methods. In practice, MAPEL varies the axes weights until multiple clinical response distributions are in agreement with published trial statistics.
Figure 3
Figure 3
Workflow for generating the ensemble of virtual populations. MAPEL can be applied iteratively to create an ensemble of acceptable axes weight solutions, or an ensemble of VPops. The general workflow is shown for clarity. First, valid VPs that meet the acceptance criteria detailed in the methods were developed. In total, a cohort of 1,206 VPs was created by introducing diversity along 51 mechanistic axes using a genetic algorithm and screening for VPs with realistic pathophysiology at baseline and feasible responses to therapy. The cohort of VPs was then used with the aggregated clinical trial data to inform the MAPEL algorithm. Valid axes weight solutions that defined VPops were randomized and used as starting points for additional iterations to create alternate virtual populations. Ultimately, 768 alternate VPops were developed for subsequent analyses.
Figure 4
Figure 4
Calibration result for a single virtual population. As described in the methods, each VPop is assessed for agreement with mean, standard deviation, and binned distribution results from clinical trial data on the basis of ACR scores. (A) The mean (dot) and standard deviation (error bar) of low-dose anti-TNF, high-dose anti-TNF therapies (infliximab, adalimumab), rituximab, and tocilizumab was calibrated to clinical trial results. (B) The ACR distribution was also calibrated to clinical trial data. The y-axis indicates the fraction in the bin. Note that every virtual population had to compare well against these criteria, with a composite goodness-of-fit p-value greater than 0.05, to be included in subsequent analyses.
Figure 5
Figure 5
Virtual populations were selected with composite goodness-of-fit p-values greater than 0.05. VPops developed with the MAPEL algorithm give good agreement with clinical trial results as assessed by the composite goodness-of-fit criterion. Each circle represents a VPop, and smaller p-values (x-axis) imply a worse goodness-of-fit. MAPEL results were filtered for VPops with a composite goodness-of-fit greater than 0.05 and used for further analyses (vertical grey line). Of particular interest was the response to treatment with rituximab (y-axis), and the reported mean clinical response at 6 months is shown by the heavy black line. To explore the mechanistic characteristics of populations that respond well, VPops with an ACR-N response greater than 48 were contrasted with VPops with responses less than the mean.
Figure 6
Figure 6
Alternate virtual populations exhibit distinct mechanistic patterns that distinguish populations that respond well to rituximab. (A) Mechanistic differences in the VPops were apparent from a heatmap analysis. VPops were ordered by increasing ACR-N response to rituximab. The color indicates the weight on the higher axis bin, since each axis was split into two bins for assigning weights with MAPEL. Darker red implies a higher weight on the higher-valued bin, meaning the activity governed by the axis is generally high in the VPop. (B) Differences were apparent between VPops with greater responses to rituximab and those with less than the mean response. VPops with a mean ACR-N response greater than 48 (white boxes) were contrasted with virtual populations with less than the observed clinical mean response (grey boxes). Interquartile ranges are also depicted by the box widths. The heavy black line shows the median. Mechanistic axes are ordered from the largest difference between the medians at the top of the figure to least differentiated at the bottom of the figure.
Figure 7
Figure 7
Sensitivity analysis in the ensemble of populations proposes IFNβ-mediated mechanisms that dominate the clinical response to rituximab. Sensitivity analysis was performed by altering the indicated IFNβ-effect in the biosimulation for each VP while administering rituximab, re-running simulations for the entire cohort, and recalculating population-level statistics using the previously calculated axes weight solutions from MAPEL. The populations are ordered along the x-axis by increasing response to rituximab, and the shade indicates the mean ACR-N score for the VP cohort. In the top series, one effect of IFNβ was allowed to respond to the application of rituximab. In the bottom series, one effect of IFNβ was held fixed during the response to the application of rituximab. VPops were primarily sensitive to the effects on IFNβ on macrophage IL-10 synthesis and FLS proliferation.
Figure 8
Figure 8
Biomarker analysis in virtual populations robustly identifies synovial mediators that predict response to rituximab. An exhaustive linear regression with five regressors was performed with each VPop to identify the five synovial analytes most predictive of the response to rituximab.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Firestein GS. Evolving concepts of rheumatoid arthritis. Nature. 2003;423:356–361. doi: 10.1038/nature01661. - DOI - PubMed
    1. Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, Bingham CO, Birnbaum NS, Burmester GR, Bykerk VP, Cohen MD, Combe B, Costenbader KH, Dougados M, Emery P, Ferraccioli G, Hazes JMW, Hobbs K, Huizinga TWJ, Kavanaugh A, Kay J, Kvien TK, Laing T, Mease P, Ménard HA, Moreland LW, Naden RL, Pincus T, Smolen JS, Stanislawska-Biernat E, Symmons D. et al.Rheumatoid arthritis classification criteria: an American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis & rheum. 2010;62:2569–81. doi: 10.1002/art.27584. - DOI - PubMed
    1. McInnes IB, Schett G. Cytokines in the pathogenesis of rheumatoid arthritis. Nat Rev Immunol. 2007;7:429–442. doi: 10.1038/nri2094. - DOI - PubMed
    1. Felson D, Anderson J, Boers M. The American College of Rheumatology preliminary core set of disease activity measures for rheumatoid arthritis clinical trials. Arthritis & rheum. 1993;36:729–740. doi: 10.1002/art.1780360601. - DOI - PubMed
    1. Anderson JJ, Wells G, Verhoeven AC, Felson DT. Factors predicting response to treatment in rheumatoid arthritis: the importance of disease duration. Arthritis & rheum. 2000;43:22–29. doi: 10.1002/1529-0131(200001)43:1<22::AID-ANR4>3.0.CO;2-9. - DOI - PubMed

Publication types

  NODES
admin 1
Note 1
twitter 2