Urban Borštnik and Tomaž Curk (2002) . Prešeren awards for students.

Full text not available from this repository.

Abstract

Mutations are a key tool that geneticists use in exploring biological processes. By using them, they can determine which genes play a role in some biological process and observe their mutual influence. They most often summarize their observations and conclusions about gene relations and influences in a graph called a genetic network. To build a genetic network biologists use the gentic data that includes information on mutants (changes in genotypes) and data on the resulting change in biological processes due to mutations induced (changes in phenotypes). To infer relatons between genes, biologist use reasoning (inference) patterns of the type ''IF a certain combination of experiments is present THEN we may propose a certain relation between genes and biological processes.'' Although such an analytical process is established among geneticists, it has only recently been formalized and implemented in the GenePath program. The main contribution of the project reported in this work is the development of an algorithm for proposing additional genetic experiments and an algorithm for generating hypothetical genetic networks that are consistent with genetic data. For this this, we have extended GenePath, which originally generated only one genetic network and did not suggest any additional experiments. The algorithm that proposes additional genetic experiments first detemines which pairs of genes are unrelated due to the lack of experiments. Zhe suggestions, which are determined on the basis of background knowledge in the form of inference patterns, include these missing experiments. The algorithm for generating hypothetical networks is recursive and generates all possible networks consistent with the experimental data. Constraints derived from known gene relations limit the number of generated networks, and the algorithm discontinues generating the networks that are inconsistent with the constraints. We tested both algorithms on experimental data on Dictyostelium discoideum maoebas and Caenorhabditis elegans nematodes. Both algorithms have proven to perform well in the analysis, planning of additional experiments with mutants, and gradual building of relatively large genetic networks. We therefore conclude that the methods we have developed and the tools we have implemented in this work are adequate and useful for experts in the field of functional genomics, and csn be used as ''intelligent assistants'' that aid in the analysis of genetic data and the hypothsizing on genetic networks.

Actions (login required)