Fragment dictionary

Structural keys are binary vectors in which each element is true or false and denotes presence or absence of a corresponding structural feature. Examples of such structural features are common functional groups, rings and ring systems, specific atom types, and so forth. Occurrence frequency of specific features may also be encoded by mapping each feature to a set of bits. Structural keys rely on the use of a predefined fragment dictionary, which specifies which feature is encoded by every bit or bit combination of the key. 

Two different MACCS keys (or MDL keys) [MACCS keys - MDL Information Systems Inc., 2008 Durant, Leland et al, 2002] are commonly encountered, one containing 960 bits and the other, which is public, containing a subset of 166 bits (ISIS keys). The fragment dictionary is based on a number of atom types, atom pairs, and custom atom environments. There can be a one-to-one relationship between the structural features and bits, or hashing can be used to create a many-to-one or many-to-many relationship between the features and bits. 

The fingerprint methods can be divided into dictionary-based and hashed-based methods. In the dictionary-based methods, such as the MDL MACCS keys [12] and BCI fingerprints [13], a binary fingerprint is defined in which each bit represents a particular substructural fragment contained in a fragment dictionary. The fingerprint... 

The WLN-fragment search was notable in that its use did not require full knowledge of the notation and its rules. A "reading" knowledge of WLN was all that was required, and armed with a WLN-fragment dictionary and a short training course, a few adventurous chemists actually used the system themselves. It was not for lack of enthusiasm that the system did not receive full... 

Figure 2 Simple illustration of bit string encoding of chemical structure, (a) Sample of a fragment dictionary-based approach, (b) Sample of a hashing scheme using a path-based decomposition of the structure. The asterisk denotes an element in the bit string where a collision has resulted from the hashing procedure.

Assume that a screen set is to be created that contains N screens. Partition the fragment dictionary into N partitions, each of which contains approximately the same number of fragment occurrences. The range of values encompassed within each such partition then corresponds to one of the screens that are available for assignment to database structures or to query substructures. 

The ideal partition frequency, P, is calculated. This is done by summing the frequencies of occurrence for all of the fragments in the fragment dictionary and dividing this sum by the number of partitions required, N. 

The mean partition size for each section is then calculated and the division points, i.e., the boundaries between one partition and the next, are placed as close as possible to the required positions in the complete fragment dictionary. 

Perhaps the most significant feature of this virtual screen set is the incorporation of a rare and/or theoretical screen that is set when any generated fragment is not found in the "formal part of the screen fragment dictionary. The frequency of occurrence of this rare/theoretical screen is very rare. It can be viewed as a fail safe screen since it can logically complete any set of alternative screens for a synthetic screen without significantly detmorating the search speed. 

---