Sebastian Gomez

Despite significant progress in the molecular understanding of cancer, stratification of risk in patients remains a challenge. Focus has shifted from clinical parameters to molecular markers, such as expression of specific genes and selected genomic abnormalities, in order to improve the accuracy of treatment outcome prediction. In Tamayo et al 2010, we showed how integration of high-level clinical and genomic features, in a cumulative log-odds Bayesian nomogram, could yield clinical prediction models that are more comprehensive, accurate, and biologically interpretable on a patient-by-patient basis. The proposed method evaluates the weight of evidence as the main unit of predictive potential. This measurement is used to determine the appropriateness and usefulness of a feature in predicting the correct clinical outcome. After the feature selection stage, the model computes a cumulative weight of evidence by integrating the evidence from each feature. The cumulative weight of evidence is then used to predict a patient’s clinical outcome. The model also provides a pictorial representation of the strength and consistency of the feature’s evidence on a patient-by-patient basis. In this paper we explore the addition of pairs of features to the previous model. The pairs are selected based on their joint conditional evidence with respect to the relevant clinical prediction. Our results demonstrate an improvement in the prediction of platinum resistance in ovarian cancer, and relapse after treatment in medulloblastoma, when compared to our previous model using “single” features. Additionally, the pairing of features serves to identify potential interactions between genomic features that will be the subject of further follow-up research.

 

PROJECT: Weight of Evidence Inference and Bayesian Nomograms using Genomic Feature Pairs

Mentor: Pablo Tamayo, Cancer Program

Presentation

Sebastian Gomez

Spending time at the Broad Institute is perhaps any researcher’s dream. Throughout the summer I had the opportunity to meet with many incredibly smart people who motivated me to challenge myself and pursue a career in science. There were many exceptional things that I learned at the Broad, but the one thing that stuck with me was that discovery and innovation is a community effort, not a single man’s task. At the Broad, you can truly see the effectiveness of this ideology