Cambridge Healthtech Institute’s sixth annual “World Pharmaceutical Congress” focused on lead optimization with the goal of increasing the efficiency of the hit-to-lead-to-drug candidate process. A roundtable discussion moderated by Christopher B. Cooper, Ph.D., associate director of early discovery chemistry at Bristol-Myers Squibb, asked the question “Lead Optimization vs. Lead Rejection: When is Enough Enough?”
With more than 20 years experience at Bristol-Myers Squibb and Pfizer, Dr. Cooper emphasized that it is an extremely competitive environment for pharma large and small.
“In developing chemical leads from HTS, we are faced with limited resources to interrogate new chemical matter and many options to pursue. When do you decide to fish or cut bait?” A formal hit-to-lead process has been established in many companies to identify high-quality leads serving as better starting points for lead optimization.
Dr. Cooper said that strategic decisions, based on initial in vitro potency/efficacy, breadth of chemotype structure-activity relationships, chemical tractability, and potential intellectual property position, for example, made early-on to focus the scope of synthetic efforts for a given chemotype may ultimately doom a program/approach due to inherent, unknown deficiencies of the original chemical starting point in this optimization effort.
Answering this fight-or-flight question, Dr. Cooper listed the elements he believes are key, beginning with safety and including intellectual property assessment and physical chemical parameters. “If you have a range of attractive chemotypes, you must be rigorous about ascertaining the safety of the chemical series. Very rarely can you optimize your way out of a major safety liability.” Selectivity information, of which toxicity is just one component, is vital to advancing a candidate.
Dr. Cooper also noted that at some stage in this whole process there is a need for the discussion of subjective versus objective decision-making criteria. Scientists could be overruled, rightly or wrongly, by top management or marketing and sales who may have different agendas. He believes the “pharmaceutical discovery industry” must be aware of and responsive to these disparate factors.
Rational Design
Rational design is a promising approach because it provides a tool that allows screening of far fewer compounds, reducing many thousands to hundreds. It is particularly useful, Dr. Cooper added, for kinases, nuclear hormone receptors, soluble enzymes, and certain ion-channel targets where one might develop structure activity relationships through structure-based design aimed at specific molecular binding sites for a given target of interest.
In a presentation on the use of rational design, “Recent Advances in Structure-Based Lead Optimization,” Woody Sherman, director of applications science at Schrodinger, presented improvements in lead optimization using the Schrodinger software suite.
“Crystal structures provide a great way to better understand how small molecules interact with the target of interest to increase binding strength or target specificity, whether with proteins, DNA, RNA, or glycoproteins. When crystal structures are not available, the group can build homology models or create induced-fit models,” Sherman noted.
Schrodinger’s Induced-fit Docking (IFD) methodology was developed to address situations in which it is important to account for protein flexibility and provides a computational surrogate to generate quite accurate models of protein-ligand complexes, he said. The iterative procedure combines flexible ligand docking and protein structure refinement.
“Even in cases of docking similar molecules, accounting for induced-fit can be important in generating the correct protein-ligand complex structure. Without the right complex structure, one cannot hope to accurately predict ligand-binding affinities,” Sherman observed.
The quality of the results obtained in an iterative protocol like IFD is highly dependent on the quality of the programs used, Sherman said, with factors such as the force field and sampling algorithm playing important roles. “It is not enough to just combine any programs for docking and protein refinement without performing the right validation. It is important to get the science right and combine methods in an intelligent way,” said Sherman.
“Once we obtain an accurate receptor-ligand complex, we use Prime MM-GBSA to rank order the binding of molecules within a congeneric series, as is necessary in lead-optimization projects.”
The software emphasizes using an accurate forcefield and solvation model. Additionally, the method now can use quantum mechanical charges directly in the calculation. Finally, Sherman discussed the use of MCPRO+, released last month, which uses Monte Carlo statistical mechanics simulations to compute free-energy changes between molecules via Free Energy Perturbation (FEP) calculations.
“MCPRO was developed by Bill Jorgensen at Yale,” Sherman said. “Schrodinger has kept the rigorous scientific methods in place while simplifying the interface to make FEP calculations accessible to nonexperts. The integration of these tools and other advances in the Schrodinger software suite is making lead optimization by computational approaches more feasible and successful. We are seeing successes more frequently in real drug discovery projects.”
Lead Selection
Genedata’s Joe Shambaugh presented on “Optimizing the Lead Identification Process by Integrating Experimental and In-Silico Data on Hit Compounds and their Interactive Selection by Cross-Functional Expert Teams.” Shambaugh pointed out that lead selection has moved beyond simple activity and selectivity considerations and has become “a truly multidimensional exercise, driven by the need to reject compounds unsuitable for drug development as early as possible and to identify compounds with the highest optimization potential early.”
Bringing together all available information and people with relevant expertise to this process is still a major challenge in the industry. Many times researchers use Excel with information captured on multiple spreadsheets, over time, over many users. Enterprise solutions exist, Shambaugh said, but are not well accepted and frequently result in “work-arounds.”
The Genedata solution, Hit Profiler, provides the means for users to evaluate compounds based on measures such as IC50 values, as well as to enrich the list of all screening hits with information from their area of expertise and other corporate sources, according to Shambaugh.
The combined information is used to finally prioritize compounds to progress. Compared to a simple, sequential filtering process, this approach ensures well-balanced results and avoids information loss in the handover, he noted.
Shambaugh pointed out that the Genedata software, while in a familiar spreadsheet format, features an “application programming interface that is flexible and easy to use.” For example, it allows implementation against any corporate data repository, he said.
Designed for use by the entire discovery team, privileges can be assigned to ensure security. Legacy data can also be integrated, adding an assessment of how a compound performed in previous screens.
Information Challenge
In his presentation, “Aligning Target Selection with Compound Design—Leveraging Prior Knowledge for Efficient Discovery,” John Overington, Ph.D., senior director of discovery informatics at BioFocus DPI, gave his perspective on what he describes as the information challenge. “In 1986 it was possible to store all the data you really needed in your head. Since then many drug discovery technologies have emerged—genomics technologies are producing more known sequences; there are a plethora of compounds coming from combinatorial chemistry; and HTS campaigns produce huge volumes of data.
“This technological advancement, allied to big budgets, has produced vast amounts of data. Today, we recognize the need to manage data more efficiently as well as the challenge posed by doing so.”
BioFocus DPI, a Galapagos company, launched its latest discovery knowledge-base, Kinase SARfari, at the “World Pharmaceutical Congress.” This fully integrated data repository and research workbench focuses on the protein kinase family of drug targets, explained Dr. Ovington. “The tool combines chemical and biological data from internal proprietary and public sources in a single, dynamic, and responsive system.”
Kinase SARfari enables the user to design compounds and focused libraries against specific protein kinases and rapidly optimize compound discovery through exploration of structure-activity relationships, Dr. Ovington noted. Through effective integration and mapping of data from diverse biological and chemical sources, Kinase SARfari increases productivity, he continued.
The system comes prepopulated with all human protein kinase sequences and a large collection of model organism orthologs; protein kinase clinical candidates and FDA-approved drugs; more than 700 3-D structural domains from the Protein Data Bank with complete binding-site-based structural superposition; and many thousands of different compounds and screening data taken from the primary literature; and, finally, five binding-site definitions based on analysis of known ligand-binding footprints with precalculated binding-site physicochemical distances. On top of this seed data, users can readily add their own proprietary internal data for fully integrated analysis, Dr. Ovington added.
“It’s quite easy to underestimate the software engineering task,” Dr. Overington concluded. “Without new data and care, databases just wither on the vine. Our goal is to build long-lasting, well-maintained solutions.”
Knowledge-based Lead Generation
In “Knowledge-Based Lead Generation and Optimization,” Zhengming Chen, Ph.D., senior director of chemistry at DOV Pharmaceutical, discussed the discovery of NMEs to treat neuropsychiatric disorders. Knowledge-based lead-generation strategy, Dr. Chen observed, includes the design of mimics of the endogenous ligands/mediators, the identification and use of privileged structures, and analog research from known ligands and drugs.
The objective of DOV’s monoamine uptake inhibitor program is to discover and develop novel structures to gain new patent positions with a balanced triple uptake inhibition profile or inhibition of at least two monoamine transporters. The program explores different selectivity profiles among monoamine transporters and aims to improve therapeutic activity by modification of the interaction of secondary targets. DOV chemistry strategy utilizes knowledge-based lead generation and optimization, bioisostere evolutions, chiral switch, and ligand-based drug design.
Dr. Chen observed that enantiomers, metabolites, combinations, and repositions of existing drugs are valid sources for new, patentable NMEs, and that mimics of endogenous ligands/mediators, privileged structures, known ligands, and analog research can all yield useful leads.
Structure-target relationships have been explored at DOV to determine what chemical structural features are required for a specific target protein such as a monoamine transporter. Dr. Chen’s presentation specified two features.
The first is the presence of a basic amino group in the compound; all 25 marketed transporter inhibitors have this structural feature as either a primary, secondary, or tertiary amino group. The second key feature is the presence of a phenyl group—18 of 25 transporter based drugs have two phenyl groups; six have one phenyl group plus a hydrophobic ring or chain; and one has two aryl groups.
DOV’s monoamine uptake inhibitor program now has two development candidates for the treatment of depression or alcoholism in Phase Ib/II and Phase I trials. A third NME is at the candidate selection stage, with several others in the lead-optimization stage.