Survival analyses are essential in biomedical research for understanding the time until an event of interest, such as organ failure or patient death. To identify risk factors for kidney allograft failure (GF) related to HLA amino acid-level mismatches (AA-MM), ‘Feature Inclusion Bin Evolver for Risk Stratification’ (FIBERS), an evolutionary binning algorithm for modeling and feature learning, was developed.

Previous versions of FIBERS optimized bins of AA-MM features based on their ability to stratify donor-recipient pairs into two kidney GF risk groups (low and high) through a single threshold value per bin. This project focused on expanding FIBERS to automatically optimize the number of risk groups for bins by allowing FIBERS to stochastically search the threshold space for multi-group bins (3 risk groups).

The new version of FIBERS includes updated genetic operators, new bin fitness evaluation metrics, and reworked bin threshold representation. The algorithm was tested on simulated survival data, and next steps include running a full simulation study with with 30 replicates to discover the algorithm's strengths and weaknesses.