Modeling analysis of differentiation and proliferation profiles of myeloma clonotypic fractions. (A) Likelihood analysis: the different cell types are represented by the gray squares and transitions between them represented by arrows (with associated rate parameters; see supplemental Methods). The set of black solid arrows indicate the transitions in the null model that assumes linear transition from CD19⁺ cells to PCs via Pre-PC and CD138^low cell types. Each dotted arrow indicates an included transition tested (ie, Pre-PC → CD19⁺ cell, CD138^lowPC → CD19⁺ cell, PC → CD19⁺ cell transitions) with respect to the null model using a likelihood ratio test. Black dotted lines indicate transitions that were not significant. The PC to Pre-PC transition (indicated by the dotted line) showed a P value of .031, indicating significance at the 5% level. (B) Bayesian analysis of the PC to Pre-PC transition: to further investigate this transition, an MCMC algorithm was developed to fit a fully Bayesian model. The box plot shows the marginal likelihood for the null model and for the null model, including the PC to Pre-PC transition for 10 runs of the MCMC algorithm with different starting values. The Bayes factor of 5.06, here calculated as the ratio of the marginal likelihoods, represents strong evidence for the inclusion of the PC to Pre-PC transition.