Metabolism can be regarded as a network of chemical reactions catalyzed by enzymes, In this scenario, a coordinated sequence of biochemical reactions is called a metabolic pathway. A metabolic pathway interacts with others leading to junctions where they meet or cross.
Prediction and reconstruction of metabolic pathways from molecular data is a fundamental way of understanding the logic of life.
In particular, the reconstruction of metabolic pathways could help to identify novel genes. A metabolic pathway can also be reconstructed to accomplish a given metabolic function. These reconstructed pathways are very useful in the study of early steps of the conception and design of a process where a pathway must be chosen for the production of the desired product.
Therefore, modeling biochemical systems on computers results in some important benefits, such as: low cost, the possibility of predicting molecules, which are hard to be experimentally obtained, the speed of computational experiments, which are faster than lab ones, and, very important in this global-warning time, the safeness of the environment.
On the other hand, an artificial chemistry is a tool for computationally modeling biochemical systems, such as metabolic pathways.
It is aimed at answering qualitative rather than quantitative questions about the system, and consists of ob jects, rules, and the definition of the dynamics of the system. Unfortunately, an artificial chemistry defined by a set of molecules and a set of chemical reactions is often exponential in the size of these sets. Thus, artificial chemistries are only known for very small instances, i.e. just involving a few dozens of molecules and chemical reactions.
This thesis is aimed at reconstructing artificial chemistries modeling metabolic pathways that can accomplish a given metabolic function of transforming molecules into others by using a set of biochemical reactions. Here, we adopt an artificial chemistry model based on relabeling graph grammars and whose definition of the dynamics still needed to be established. Therefore, we introduce two new methods to establish such a definition. In our first method, based on unidirectional chemical search, we impose constraints related to complexity considerations. The metabolic pathways obtained show that most of them lie outside the artificial chemistry. This was the reason for the development of a second method, which is based on bidirectional chemical search. In this second method, we impose more constraints related to complexity considerations and add constraints related to biological meaningfulness.
Furthermore, this model requires an atom mapping while characterizing chemical reactions, which are represented by chemical reaction graphs. Thus, we also deal with the computation of an atom mapping in this thesis. Molecules at the beginning of a chemical reaction are called substrates whereas molecules at the end are called products. An atom mapping is a bijection from substrate to product atoms. On the other hand, the classical principle of minimum structure change states that a chemical reaction normally occurs through the formation and breaking of the least number of bonds while transforming substrates into products. Therefore, an optimal atom mapping is an atom mapping modeling this principle.
Here, we present a new method to compute an optimal atom mapping of a given chemical reaction. Our method exploits the knowledge of the atomic rearrangement pattern within molecules involved in a chemical reaction. Thus, the computation of an optimal atom mapping is performed by a series of chemical substructure searches between the substrate and the product of the given reaction. A classification according to these patterns is widely used by chemists and we have extended this classification.
Atom mappings are useful, for instance, for drug design, consistency checking of metabolic pathway databases, and classification of reactions.
Metabolic pathways are often represented as directed hypergraphs, with substrates and products as nodes and biochemical reactions as hyperarcs or, alternatively, as bipartite graphs with substrates and products as one node class, and biochemical reactions as the other node class. Therefore, this artificial chemistry is represented as a directed graph, where chemical graphs, denoting substrates and products, are vertices and applications of chemical reaction graphs are arcs. In particular, this graph is formed by optimal pathways, which are series of biochemical reactions with the least possible total number of broken and created bonds.
We have focused on a metabolic pathway database, and we have applied our method based on bidirectional chemical search to some pathways stored in this database. As a result, our method have revealed thousands of metabolic pathways, most of them biologically meaningful, and in particular, various novel metabolic pathways.
Moreover, we have applied our method for computing an optimal mapping to a substantial portion of this database. As a result, our method has found an optimal mapping for a large number of chemical reactions stored in this database. We have also ensured consistency of a large number of biochemical reactions, and have identified several inconsistencies in the information about chemical compounds and biochemical reactions contained in this database. Furthermore, we have also developed a tool, which given a biochemical reaction and its atom mapping produces a pair of GIF files containing a diagram of the reaction with appropriate atom coloring to illustrate the correspondence among substrate and product atoms. This tool relies on an algorithm for the graphical depiction of chemical structures.
Thus, our methods constitute new tools for biologists and biochemists while analyzing and synthesizing metabolic pathways. On the other hand, by ensuring the consistency of metabolic pathways, we contribute to the advancement of research areas like systems biology since information stored in these databases is one of the main sources in this area, thus it needs to be correct.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados