Computational Discovery of Phenotype Related Biochemical Processes for Engineering

Application of bioengineering technologies for enhanced biological hydrogen production is a promising approach that may play a vital role in sustainable energy. Due to the ability of several naturally occurring microorganisms to generate hydrogen through varying metabolic processes, biological hydrogen has become an attractive alternative energy and fuel source. One area of particular interest is the production of biological hydrogen in organically-rich engineered systems, such as those associated with waste treatment. Despite the potential for high energy yields, hydrogen yields generated by bacteria in waste systems are often limited due to a focus on microbial utilization of organic material towards cellular growth rather than production of biogas. To address this concern and to improve upon current technological applications, metabolic engineering approaches may be applied to known hydrogen producing organisms. However, to successfully modify metabolic pathways, full understanding of metabolic networks involved in expression of microbial traits in hydrogen producing organisms is necessary. Because microbial communities associated with hydrogen production are capable of exhibiting a number of phenotypes, attempts to apply metabolic engineering concepts have been restricted due to limited information regarding complex metabolic processes and regulatory networks involved in expression of microbial traits associated with biohydrogen production. To bridge this gap, this dissertation focuses on identification of phenotype-related biochemical processes within sets of phenotype-expressing organisms. Specifically, through co-development and application of evolutionary genome-scale phenotype-centric comparative network analysis tools, metabolic and cellular components related to three phenotypes (i.e., dark fermentative, hydrogen production and acid tolerance) were identified. The computational tools employed for the systematic elucidation of key phenotype-related genes and subsystems consisted of two complementary methods. The first method, the Network Instance-Based Biased Subgraph Search (NIBBS) algorithm, identified phenotype-related metabolic genes and subsystems through comparative analysis of multiple genome-scale metabolic networks. The second method was the multiple alignments of metabolic pathways for identification of conserved metabolic sub-systems in small sets of phenotype-expressing microorganisms. For both methodologies, key metabolic genes and sub-systems that are likely to be related to hydrogen production and acid-tolerance were identified and hypotheses regarding their role in phenotype expression were generated. In addition, analysis of hydrogen producing enzymes generated by NIBBS revealed the potential interplay, or cross-talk, between metabolic pathways. To identify phenotype-related subnetworks, three complementary approaches were applied to individual, and sets of phenotype-expressing microorganisms. In the first method, the Dense ENriched Subgraph Enumeration (DENSE) algorithm, partial prior knowledge about the proteins involved in phenotype-related processes are utilized to identify dense, enriched sets of known phenotype-related proteins in Clostridium acetobutylicum. The second approach utilized a bi-clustering algorithm to identify phenotype-related functional association modules associated with metabolic controls of phenotype-related pathways. Last, through comparison of hundreds of genome-scale networks of functionally associated proteins, the á, â-motifs approach, was applied to identify phenotype-related subsystems. Application of methodologies for identification of subnetworks resulted in detection of regulatory proteins, transporters, and signaling proteins predicted to be related to phenotype-expression. Through analysis of protein interactions, clues to the functional roles and associations of previously uncharacterized proteins were identified (DENSE) and hypotheses regarding potentially important acid-tolerant mechanisms were generated (á, â-motifs). Similar to the NIBBS algorithm, analysis of functional modules predicted by the bi-clustering algorithm suggest cross-talk is occurring between pathways associated with hydrogen production. The ability of these phenotype-centric comparative network analysis tools to identify both known and potentially new biochemical process is important for providing further understanding and insights into metabolic networks and system controls involved in the expression of microbial traits. In particular, identification of phenotype-related metabolic components through a systems approach provides the underlying foundation for the development of improved bioengineering technologies and experimental design for enhanced biological hydrogen production