American Statistical Association
When two tumors arise in a patient, a key question is to determine whether they are independent primary tumors, or whether one is a metastasis from the other. This can be done by assessing the mutations characterizing each tumor, and comparing the shared ones (favoring clonal relatedness) to the non-shared ones (favoring independence). We have previously developed a random-effect model estimating the key parameters needed to derive individual probabilities of clonal relatedness. However, while the maximum likelihood estimate gave unbiased results in most of our settings, we noticed specific cases where the estimates were not acceptable, notably giving a proportion of clonal cases exactly equals to 0 or 1. These values lead, in turn, to individual probabilities of being clonal equal to 0 or 1 for all cases. Such problems arise mainly when the information is scarce: few cases or few mutations observed.
To obtain more plausible estimates in those settings, we investigated Bayesian estimations of our model. Using log-normal prior distribution for the proportion of clonal cases, the approach naturally excludes 0 and 1 from the acceptable solutions. We compared the performance of the Bayesian approach to the maximum likelihood estimator using a simulation study, and the advantages of both approaches are discussed. The difference is illustrated on a breast cancer dataset.
|Date:||Wednesday, April 11, 2018|
|Time:||4:00 - 5:00 P.M.|
Memorial Sloan Kettering Cancer Center
Department of Epidemiology and Biostatistics
485 Lexington Avenue
(Between 46th & 47th Streets)
2nd Floor, Conference Room B
New York, New York
**Outside visitors please email firstname.lastname@example.org for building access. You must be on the security list to enter the floor.