Speakers: Brad Olson, PhD
Brad Olson, PhD
Kansas State University
Bradley Olson is an Associate Professor at Kansas State University where he studies how multicellular organisms evolve. He received his PhD in 2008 from Michigan State University and was a Ruth L. Kirschstein Post-doctoral Fellow at the Salk Institute. At Kansas State, he is the Associate Director of Bioinformatics and the PI of two current NSF funded projects that use bioinformatics and genomics approaches toward understanding multicellular evolution.
Predictive Modeling of Multicellular Evolution
The evolution of multicellular organisms is a major transition of biological complexity that has occurred multiple times on Earth. However, the theoretical and molecular basis for why multicellular organisms have evolved multiple times on this planet is poorly understood. Using large scale “-omics” data sets from the volvocine algae, who recently evolved multicellularity, we are developing predictive models for genes and pathways that are important for evolving multicellular complexity. Model development requires an end-to-end solution that presents multiple challenges to model development. Unsupervised feature engineering of a high dimensional data is the biggest challenge to model development. Current model development is focused on a combining k-nearest neighbor clusters combined with multilayered convolutional neural networks. These layers are then fed into Bayesian neural network for predictions. Models are then trained with “-omics” data from wild type unicellular and multicellular organisms as well as similar data obtained from unicellular mutants of undifferentiated multicellular organisms. This talk will focus on overcoming the challenges of model development with highly dimension data with low replication.