TY - JOUR T1 - Inferring restrictions in the temporal order of mutations during tumor progression: effects of passenger mutations, evolutionary models, and sampling JF - bioRxiv DO - 10.1101/005587 SP - 005587 AU - Ramon Diaz-Uriarte Y1 - 2014/01/01 UR - http://biorxiv.org/content/early/2014/06/12/005587.abstract N2 - In cancer progression, fixation of some driver mutations (those causally involved in the disease) can depend on the presence of other drivers. The majority of mutations present in cancer cells, however, are not drivers but passenger mutations that are not involved in disease progression.Several methods have been developed to identify restrictions in the temporal order of accumulation of driver mutations from cross-sectional data, but the few available comparisons of performance assume that drivers are known, and have not examined the effects of sampling. Here I conduct a comprehensive comparison of the performance of all available methods.In contrast to previous work, I embed order restrictions within evolutionary models of tumor progression that include passengers and drivers. This allows me to asses the effects of having to filter out passengers, of sampling schemes, and of deviations from order restrictions.Poor choices of method, filtering, and sampling can lead to large errors in all performance metrics. However, one filtering approach emerges as a reasonable compromise when drivers are unknown. The best method for reconstructing order restrictions are Oncogenetic Trees: over a range of scenarios they do a very good job, superior to Conjunctive Bayesian Networks and Progression Networks. Single cell sampling provides no advantage, but sampling in the final stages of the disease vs. sampling at different stages can have severe effects depending on the evolutionary model. Evolutionary model and deviations from order restrictions can have major, and counterintuitive, interactions with other factors that affect performance.This paper provides practical recommendations for using these methods with experimental data. Moreover, it implements a framework that shows that it is both possible and necessary to embed assumptions about order restrictions and the nature of driver status within evolutionary models of cancer progression to evaluate the performance of inferential approaches. ER -