In moving from hit to lead or from lead to clinical candidate, a very well-known medicinal chemistry technique is to change a portion of the chemical structure of one of the promising molecules in the project. Sometimes this fragment to be changed is on the periphery of the original molecule, in which case the substitutions generate a small library of analog molecules. At other times, particularly when less conservative changes are desired, a central portion of the molecule may be replaced. This is often referred to as “scaffold-hopping” and may demand completely new approaches to synthesis. BROOD is facile at assisting modelers and chemists with both of these fragment replacement tasks.
Bioisosteric replacement is a very well-known and the most conservative form of the general fragment replacement process. It consists of replacing one fragment in a bioactive molecule with another fragment that is known to closely mimic the original fragment or moiety. Bioisosteric replacement allows a chemist to modulate the physical, chemical and biological properties of a small molecule while maintaining or improving its primary binding activity. BROOD is capable of identifying very conservative bioisosteric replacements for a query fragment, but BROOD also identifies more interesting fragments that are also similar to the original molecular fragment, but would not be considered bioisosteres.
Every medicinal chemist is fluent in the practice of examining a hit or lead molecule and imagining fragment replacements that might improve its properties. BROOD is a program to help chemists and modelers be more thorough in developing these ideas. BROOD uses the shape and attachment geometry of the query fragment to identify a family of similar fragments. These fragments are then ranked according to the sum of their shape and chemical similarity to the query. In some instances, a user may want a very conservative change and will consider those most similar fragments. In other cases, a more drastic change may be in order and fragments that have the same shape and attachments, but different chemistry may be of greater interest. In either case, BROOD can be helpful in identifying interesting fragment replacements for compound optimization.
Chemists and modelers are not usually directly interested in the fragments that can conservatively replace the fragment in their original molecule, but instead, they are more interested in the analog molecules generated by swapping new fragments for the original fragment. BROOD identifies fragments that can be used to replace a user-defined portion of a lead molecule. BROOD then builds the new fragments back into the original lead molecule so that the output is the analog molecule of the original molecule. In generating these analogs, the conformations and orientation are built and minimized with MMFF in order to give the most similar shape to the input conformer of the original ligand. This process allows users to visually inspect the similarity of new analogs and efficiently use them in down-stream applications.
In many projects mature enough to have a hit or lead, one or more protein-ligand co-crystal structures are available. BROOD is able to take these protein structures as input and utilize to protein structure to eliminate replacement fragments which will no longer allow the analog to fit in the binding pocket. As stated above, when BROOD builds the analog molecules, it generated conformers and orientations that are as similar as possible to the input molecule. If the input ligand for BROOD is pulled from a cocrystal structure, BROOD’s hitlist of analogs will already be oriented in the active site. If the protein is also given as input, BROOD will check for ligand-protein close-contacts and eliminate analogs which clash with the protein. When examining analogs built by BROOD, the protein crystal structure can be viewed to assess potential protein-analog interactions. The BROOD hitlist visualization can display the protein surface as well as protein-ligand hydrogen bonds.
In addition to structural information, mature projects often have well-understood SAR requirements. After selecting what portion of a molecule to replace using BROOD, a user can set constraints for required interactions. For instance, if a modeler knows there is a requirement for an acceptor in a specific region, a constraint can be assigned in BROOD’s visual interface and all the fragments identified by BROOD will contain an acceptor in the appropriate region of space.
BROOD calculates several physical properties of each potential analog and allows users to design the physical properties of acceptable molecules. At the stage of a project in which BROOD will be used, while conserving potency is desirable, additional physical properties often are of even higher concern. BROOD’s visual interface allows users to define the physical property space from which proposed analogs can be drawn. BROOD begins by assessing the physical properties of the lead molecule and setting default windows for each physical property around the values of the lead molecule (skewed towards smaller, more rigid, less hydrophobic analogs). Users can further adjust these windows to focus on the types of analog molecules they would like to explore. Calculated physical properties include molecular weight, logP, Lipinski violations, and flexibility as well as several common measures of bioavailability [Wang-1997], [Lipinski-1997], [Clark-1999], [Egan-2000], [Ertl-2000], [Veber-2002], [Martin-2005].
Not only are physical properties a major concern, but when proposing new molecules, even close analogs of lead molecules, synthetic accessibility is of critical importance. BROOD calculates a synthetic accessibility for every analog molecule. It was recently demonstrated that a historic measure of molecular complexity is as good at predicting the average synthetic-difficulty (as assessed by a group of chemists) as individual chemists are at predicting the average synthetic-difficulty [Boda-2007]. We have implemented this-type of synthetic-accessibility measure in BROOD and use it rank the hitlist, particular to sort analogs whose similarity to the lead molecule don’t differ sufficiently to distinguish between them.
One critically limiting factor for BROOD and competing programs are the databases of fragments that they search. While it is very easy to find nice analogs of simple molecular pieces, the chemical space of fragments with 9-15 heavy atoms and 2 or 3 attachment points (often core fragments) is quite large. BROOD provides a default database of approximately 6 million fragments of previously synthesized molecules covering 1-15 heavy atoms and 1-3 attachment points. We believe that this 6 million fragment database represents the best optimization of fragment coverage and database size. These fragments are organized to allow very fast search times. We have discovered that often the most interesting result are generated from replacing fragments with more than 10 heavy atoms and 2-3 attachment points. We feel that while smaller databases are sufficient for small query fragments or few attachment points, this larger database provides exciting results for the most interesting queries and provides even better results for the easier queries.
While BROOD’s database is extensive and carefully selected to be medicinally relevant, we understand that many customers will be anxious to supplement the default database with fragments generated from their own proprietary molecule collections. BROOD comes with a second program, called CHOMP, which allows users to fragment molecules, filter the fragments, generate 3D conformations, organize and index the fragments for rapid searching, and write a BROOD database in a single step. This process can be carried out from start-to-finish using a BROOD license and does not require access to OMEGA, FILTER, MolProp, or OEChem licenses. As a final additional utility, the BROOD distribution includes a utility that allows users to merge databases, so they can create a single database from OpenEye’s default database and their own corporate database. For more information, see the section on Database Preparation.