Description:
<jats:title>Abstract</jats:title>
<jats:sec>
<jats:title>Motivation</jats:title>
<jats:p>Building calibrated and discriminating predictive models can be developed through the direct optimization of model performance metrics with combinatorial search algorithms. Often, predictive algorithms are desired in clinical settings to identify patients that may be high and low risk. However, due to the large combinatorial search space, these algorithms are slow and do not guarantee the global optimality of their selection.</jats:p>
</jats:sec>
<jats:sec>
<jats:title>Results</jats:title>
<jats:p>Here, we present a novel and quick maximum likelihood-based feature selection algorithm, named GameRank. The method is implemented into an R package composed of additional functions to build calibrated and discriminative predictive models.</jats:p>
</jats:sec>
<jats:sec>
<jats:title>Availability and implementation</jats:title>
<jats:p>GameRank is available at https://github.com/Genentech/GameRank and released under the MIT License.</jats:p>
</jats:sec>