Several high-quality serum-free media formulations were discovered which allows for further, more principled, experiments to be made to accompany and expand on the discoveries made in this study. Further work should be performed on correlating biomarkers and morphological attributes to cell differentiation and proliferation, both to improve the robustness of predictions and to simultaneously optimize proliferation and differentiation. Even without these improvements, this work is still relevant to those interested in quickly optimizing their media formulations, generally in the serum-free case, and particularly in the case of difficult-to-measure objectives such as long-term cell growth. The first part of this thesis was comprised of the development of a radial basis function genetic algorithm sequential DOE scheme. It drew heavily on the work of , where a sequential DOE technique was developed on the principle of local random search in areas of high performing media. This algorithm was also dynamic by converging on high performing results and selectively searching the design space when good results were not forthcoming. Additionally, previous work in our lab provided the framework for a sequential DOE based on a truncated GA. This modified GA incorporates uncertainty in the optimal samples found by halting algorithm convergence proportional to the amount of clustering around an optima the GA finds.

By hybridizing these two methods,growing blueberries in containers a DOE algorithm called NNGA-DYCORS was developed that solved various computational optimization problems better than either method alone. It was used to optimize a 30-dimensional media for serum-containing C2C12 cell culture with the metric of growth being AlamarBlue reduction after 48 hrs of growth in 96 well plates . While it was successful at finding media that maximized this metric , the 48 hr growth metric did not generalize well to multiple passages, and the best medium found degraded over time relative to the control. To fix this underlying problem, multiple passages needed to be incorporated into the DOE process. This is a very time-consuming process as each passage takes multiple days, many more physical manipulations than simple chemical assays which introduces opportunities for contamination, and difficulty for manual experimentation. To solve this, chemical assays were supplemented with small amounts of manual multi-passage cell counts in a multi-information source Bayesian GP model which was used to successfully optimize a 14-dimensional serum-containing media for C2C12 cells . Due to the presence of multi-passage data, the final optimal medium grew cells robustly over four passages, provided nearly twice the number of cells at the end of each passage relative to the DMEM + 10% FBS control and traditional DOE method, and did so at nearly the same cost in terms of media components. In the final chapter the multi-information source GP model was extended to optimize a 26-dimensional serum-free media based on the Essential 8 media using a multi-objective metric that improves cell growth while minimizing medium cost.

Using this Bayesian metric, a broad set of media samples along the trade-off curve of media quality and cost were found, showing that a designer can be given options in media optimization. In particular, one medium resulted in higher growth over five passages while the control and Essential 8 lagged. There are several avenues for future work. First, improving the quality and robustness of the data collected. While the Passage 2 metric did generalize to additional passages , it did not do so for all media . Rather than collect additional long-term cell counts, future researchers should attempt to find biomarker correlates with long-term growth such as Pax7, MyoD, and Myogenin. Due to the lack of expertise and resources, in-depth knowledge of the cascade of signals and molecular interactions in cells were not used to their fullest extent in this thesis, and should be considered in future. Model accuracy could also be improved by incorporating additional data such as bright field image counts of cell number, fluorescent image counts of nuclei using Hoecht stains, and growth curve data . As long as the image segmentation parameters are properly tuned, additional data points like the rate of change in cell number, final cell number, cell count at each time-point, and initial cell number could be used to detect high-quality media and elucidate intra-assay correlations. While meta bolomic, genomic, and lipidomic analysis would be time-consuming to conduct given the amount of experiments we have done here, creating multi-domain models of cellular systems may be yet another route to optimization.

Techno-economic analysis of growth and cost conditions may also allow future DOE studies to translate information collected at the lab-scale closer to the trade-offs considered in industry and large volume bioreactors. Rather than using multi-objective acquisition functions such as hypervolume or desirability maximization, future DOEs could feed raw data into a computer or algebraic techno-economic model and use composite Bayesian optimization to solve the experimental optimization problem. Secondly, fundamental “white-box” studies that focus on non-DOE aspects of the system must be completed in order to constrain the complexity of future DOE studies and set up more interesting or profitable design spaces. Studying the meta bolomics of the cell lines would be very useful in defining the upper / lower bounds and important factors of the system. By knowing the limiting factors of the media using spend-media analysis, upper ranges may be adjusted and by knowing waste-product profile, potential synthetic routes may be closed. By finding that, for example, glutamine, is a limiting amino acid in the cell culture system, a higher upper bound may be set which unlocks entirely new designs . In this thesis, the genetic and morphological profile of our cell lines was not considered. In fact, the adaptation to serum-free and Matrigel-free conditions may have appreciably changed these factors to the point that the optimal media found may not generalize to most C2C12 cells. Therefore,square pots tracking the prevalence of myotubes during differentiation or staining with α-actin or myosin heavy chain antibody would both act as a double-check for the way media optimization changes the cell line and as additional data points to consider during DOE campaigns. For example, one may optimize cell growth but with a probabilistic constraint that learns which regions of the design space result in low levels of α-actin or myosin heavy chain response and chooses experiments unlikely to violate that constraint. Future work must also go beyond C2C12s and consider cells that are relevant for cellular agriculture such as bovine, porcine, or avian. As we have seen in chapter 5, media designed for C2C12 cells does not always extend to muscle cells of different animal lineages. These new lines must be adapted to serum-free conditions which would open up the design space to more industrially relevant cell lines. All of these ideas would constitute entire projects and require their own feasibility studies, but would build upon the advances made in this dissertation and body of work.Across the United States, state and regional water quality agencies are increasingly forced to take action to control non-point source contamination from agriculture. California, often at the forefront of implementing policies to protect the environment, is in an especially dire situation and is ramping up efforts at pollution mitigation.

Examples of statewide and regional endeavors include the University of California Center for Watershed Sciences’ report to the California Legislature on nitrate in drinking water , the Central Valley Regional Water Quality Control Board’s Irrigated Lands Regulatory Program, the Central Valley SALTS program, the Climate Action Reserve’s nitrogen reduction protocol, and the Central Coast Regional Water Quality Control Board’s renewal process for the Conditional Agricultural Waiver . The latter is the focus of this dissertation. California’s agricultural pollution has been exacerbated by the most severe drought on record followed by El Niño rains as well as historically lax agricultural water quality regulations. The state’s unprecedented four-year drought has led to serious water supply and water pollution problems. There is substantially less water: The state has lost roughly 11 trillion gallons from the drought , resulting in the literal sinking of farmland . And there is even less clean water: More than half of all waters have some degree of contamination , and between 2006 and 2010, rivers, streams and lakes in California saw a 170% increase in toxicity . Agriculture is the largest contributing source of water pollution in the state. New studies are showing that the unabating drought is increasing the concentration of pollutants in water resources , and that while the predicted 2016 El Niño is expected to bring drought improvement, it is likely to exacerbate water pollution issues by increasing water runoff and accompanying contaminants . The brunt of health problems related to agricultural water pollution has fallen on the most vulnerable, marginalized populations, with nutrients and pesticides being the primary constituents of concern. In California, over 2 million people, mostly lowincome, minority farm workers are at risk of drinking nitrate-contaminated water due in large part to agricultural pollution . Schools in Central Valley’s farmland have found such high concentrations of pollutants that they have cut off their drinking fountains to students. Nitrate-contaminated drinking water from agricultural fertilizers is a well-known risk factor for “blue baby syndrome,” a potentially fatal blood disorder resulting in reduced oxygen-carrying capacity of hemoglobin . Because these communities are among the poorest in the state, many lack the resources or technical capacity to maintain safe drinking water supplies . Pesticides are another major concern due to their more obscure impact on human health and water resources than their nutrient and sediment pollutant counterparts. Two organophosphate pesticides in particular, chlorpyrifos and diazinon, have been identified as sources of water column toxicity in California. Exposure to these pesticides has been linked to neurobehavioral deficiencies, ADHD, lung damage, and in utero health effects to babies . Despite their myriad threats to human and ecological health, fertilizers and pest control agents are indispensible farming tools, supporting growers’ livelihoods and the state’s agricultural economy. Nutrient fertilizers, both in naturally derived and inorganic forms, are necessary for crop growth and development . The use of inorganic nitrogen fertilizer on California farms has intensified over the past half century, from less than 200,000 tons in the 1950s to over 750,000 tons in recent years . Pesticides and other integrated pest management strategies can lower the risk of pest outbreaks and decrease the incidence of pest damage on crops. According to the California Department of Pesticide Regulation, 194 million pounds of pesticide active ingredients were applied to California farms in 2013. The ability of California’s agricultural industry to continue to thrive economically and produce food for much of the world while not polluting waterways depends on the difficult task of balancing environmental needs with other competing concerns . Such a challenge underscores the importance of burgeoning academic discussions within the fields of environmental policy, political science, environmental economics and environmental science around choosing appropriate policy instruments , whether those instruments have been implemented effectively and equitably , how regulatory institutions are evolving to meet changing needs and why particular policy goals are prioritized over others . These literatures will be discussed in more detail in the chapters that follow. Applied case study research on policy mechanisms to control California’s agricultural water pollution is well positioned to contribute valuable insights to these bodies of work. This dissertation uses mixed social scientific methodologies to investigate regulatory tools, governance structures, policy outcomes, and stakeholder participation relating to water pollution from agriculture in California’s Central Coast. Employing a rich mixed-methods approach, including document review, historical analysis, interviews, surveys, spatial analysis, and descriptive statistics, this dissertation aims to contribute to scholarly discourse on effective agricultural water quality policy. It is my hope that this research offers valuable data and policy recommendations that are of direct use to other academics, agricultural operators, and regional water quality agencies. Throughout the research process, from designing survey questions to choosing the best means to disseminate information, I collaborated with faculty, scientists, and regional agricultural and water quality networks to ensure that this work would be applicable and relevant. For example, in developing my survey chapter , I solicited feedback and received non-financial endorsements from four well-respected agricultural organizations for a survey sent out to over 1,000 growers on issues relating to water quality practices and regulations.