Control and optimization of E. coli picking process for DNA sequencing

Abstract

As part of its responsibility to the National Institutes for Health, the sequencing operation at the Broad Institute strives for cost-effective production. This thesis attempts to reduce variability in the sequencing operation's E. coli colony picking process-thereby improving efficiency-through the application of traditional operations improvement methodology. To achieve control over variability, the author first seeks to characterize the variability and identify its drivers, then to reduce the variability by manipulating the drivers, and finally to optimize productivity. The operations techniques utilized include fishbone cause-and-effect diagram, process flow diagram and organizational analysis. Several industrial statistical techniques such as control charting, linear regression, analysis of variance and designed experimentation are also heavily employed. Many factors were studied as candidate drivers of variability. Three criteria are used to discriminate among them: statistical significance, magnitude of effect on variability and controllability. The results show that one of the largest but least controllable factors is plate density, i.e., the number of colonies on a plate. Instead of attempting to control individual confounding factors in plate preparation, this thesis presents an alternative strategy for overcoming the plate density variability: introduction of a novel spotting process that allows for plate variability but still yields higher efficiency.