The quality of test cases is an important factor to estimate the confidence one can have in a component under test. This quality can be evaluated with mutation analysis: the quality of the test cases is evaluated by the proportion of seeded faults detected by the test cases. While the generation of a basic test cases set can be easy, improving its quality may require prohibitive effort. This paper focuses on the issue of automating the test optimization.A novel algorithm is proposed, adapted from genetic algorithms that is called a bacteriologic algorithm and is inspired by the biological phenomenon of evolutionary ecology. The approach is illustrated with test generation for a C# parser.