healthy individual, an effective therapy may consist of
restoring that cell variable to its normal state. In the above
example of spinal muscular atrophy, therapies that modify
the genomic instructions so as to increase the frequency
with which thee xonis include din the proteinare
currently being tested in clinical trials [14].
The remainder of this paper provides an overview of
the role of machine learning in building computational
models of cell variables, with an aim to understand the
genetic determinants of disease. We will advocate the
approach of learning to model cell variables as an
intermediate step, and explain how this benefits from
the growing availability of diverse types of data. We will
describe in detail two types of cell variable that our group
has been closely involved in modeling, and briefly
summarize how this research has impacted our under-
standing of spinal muscular atrophy, cancer, and autism
spectrum disorder. To place our approach in context, we
will review existing techniques that are used for scoring
disease risks. Also, we will describe data sets and machine
learning formulations of problems that enable data
scientists to work in this hugely important area.