Guided Structure Generation

The second generation process is the breadth-first approach. All the nodes of a given level are fully traced and then the next level is treated. In this approach, no backtracking procedure is carried out. The generation goes breadthwise and forward only. The breadth-first approach has been implemented in the current realization of our structure generation methodand is depicted in Fig. 11. 

It should be emphasized that the SS-selection procedure is rather important for guiding the generation process toward the most likely connectivities. The selection is carried out following a series of rules based 

FIGURE 10 Depth-first generation with backtracking, SS-selection, and breadth-first check. 

Another group of selection rules refers to the spectrum/structure relationship. The system of these rules forms an a priori selection. There is also an a posteriori selection, which will be further discussed. 

---

First, it is accepted that the basic elements forming the structure are atoms (vertices of graphs). Each atom has its own attributes i.e., the vertices are colored (here the term labeled may also be used) to represent these attributes. The attributes characterize different aspects and features of the atom within chemical structures. Practically, the input spectral information is transformed into these attributes, with the latter guiding the generation process toward the correct solution(s). The atom attributes may be classified as follows ... 

A.C. Stiel, S. Trowitzsch, G. Weber, M. Andresen, C. Eggehng, S.W. Hell, S. Jakobs, M.C. Wahl, 1.8 A bright-state structure of the reversibly switchable fluorescent protein Dronpa guides the generation of fast switching variants. Biochem. J. 402, 35-42 (2007)... 

SECS program improvements must occur in several areas. We hope to implement an up-to-date version of SECS at Merck, to overcome many present limitations. In addition, high-level strategies must be developed to guide analyses, especially as the reaction data base expands, in order to remove uninteresting precursors and keep to a reasonable size the total number of structures generated. While our experience may help define successful strategies, this is not the area of Merck s initial interest. 

More recently, another class of de novo design methods has been developed that uses medicinal chemistry transformation rules to generate new structures based on one or more input compounds. Due to the fact that these transformations are based on structural modifications made in practice during the course of optinmation projects, the resulting compounds are more likely to be synthetically accessible. Furthermore, these transformation methods have been combined with MPO methods, as described above, to guide the generation of new compoimd ideas based on the simultaneous optimization of multiple properties [41]. 

---