Advances in biotechnology have led to an explosion in the amount of data that can be generated
for diagnostic purposes; examples are DNA and RNA sequencing, digital pathology, proteomics, and
multi-color flow cytometry. Many researchers feel that this "big data" era will finally bring
us closer to personalized medicine, in which each patient is given a tailored treatment rather
than an off-the-shelf generic product. But the challenges in truly harnessing big, heterogeneous
data for patient benefit are enourmous. Since just adding more data does not automatically lead
to deeper insight or better treatment decisions, robust methodologies are required to analyze and
utilize patient data in the best possible way.
Johannes Textor's project will employ a method called Causal Bayesian networks that can integrate
very diverse types of data, while clearly distinguishing factors that have a causal role from
others that are merely predictive. Once built, such networks are capable of deriving personalized
therapy outcome predictions from all available information, and can inform clinicians which
additional tests would be of greatest value to reduce the uncertainty in this prediction.
Specifically, we will build networks to attempt predicting the result of dendritic cell vaccination
for melanoma patients, a novel treatment which our lab is currently testing in a clinical trial.
The data from this trial will be directly used to inform the network building. Among the types
of data we will use are flow cytometric data and multi-color digital immunohistochemistry images
of the tumour microenvironment.