Data Mining To Optimize A Biological Manufacturing Process
Amber Biology was asked to perform a deep dive into a large body of data describing the outcomes from a complex, biological/biochemical manufacturing process being used to produce custom biomolecules. The goal of this analysis was to attempt to identify features of the manufactured biomolecules, and components of the manufacturing process itself, that might contribute to suboptimal outcomes, and which could potentially be optimized to improve the yield and quality of manufacture.
A large set of custom algorithms were implemented in Python to retrieve and filter the data from the company's manufacturing database, and to identify correlations in the set of manufacturing outcomes, with manufacturing process parameters and the attributes of the manufactured biomolecules. In a few weeks, these algorithms were able to identify some key features of the manufacturing process that could be optimized, one of which by itself, was directly responsible for the complete and immediate loss of more than 1% of the company's output, and was able to be fixed directly following the results of the analysis.