High-throughput methods are an integral tool for downstream process development to identify promising operation conditions prior to scaling up. With increasing usage of high-throughput techniques in the industry, there is an opportunity to mine and leverage the molecular information collected in these large datasets. While most of our statistical analyses and computational models typically focus on a single molecule’s behavior within a unit operation, there is a tremendous opportunity to leverage our molecule characterization dataset to further streamline downstream process development activities with predictive models that can be applied to multiple molecules. In this work, we have explored utilizing our high-throughput datasets and various machine learning models (such as random forests and deep learning neural networks) in order to construct generalized predictive models for simulating molecule behavior in cation-exchange (CEX) and anion-exchange (AEX) chromatographic operations. These models can then be used to guide molecular assessment and high-throughput experimental design, with the goal of accelerating early-stage downstream process development.

Machine learning approaches for streamlining downstream process development

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.