Integrating fragment-based drug discovery with High Throughput Screening

By Evotec

Drug Discovery in a changing environment

High throughput screening (HTS) has been the leading paradigm of small molecule drug discovery for the biopharmaceutical industry throughout the last 2 decades. Companies have invested into HTS hardware, process integration and large compound collections of frequently more than 1 million compounds. Depending on the size of the organisation and their research goals, approximately 10 to 30 HTS campaigns are run per HTS unit per year, delivering approximately, after medicinal chemistry hit-to-lead (H2L) optimisation, five to fifteen programmes with viable lead series on an annual basis (this assumes that every second HTS generates viable lead series compounds after H2L). During the 1990s significant flaws associated with an HTS approach became apparent, in particular the propensity for such campaigns to yield hits which frequently had lipophilic, non-drug like properties. In response to this, desirable lead-like and drug-like criteria for screening compounds were developed, that have served as benchmarks for library design, HTS hit selection and downstream medicinal chemistry activities. Different desirable compound profiles have been proposed and these include Lipinski’s rule-of-5 for oral absorption and Oprea’s and Rishton’s lead-like and drug-like concepts. In addition, the emphasis on early high throughput ADME/Tox profiling has provided the impetus for the development and application of in silico filters and predictive models to aid the selection of screening compounds that are likely to be free of toxicophores and display desired ADME profiles.

The industry’s response to challenges in HTS

The industry has responded to the challenge of ensuring that appropriate compounds are screened and today HTS compound collections follow more closely the ideal of screening lead-like compounds rather than drug-like or drug-unlike compounds. In parallel to this development, it has become recognised in the field that target classes with a high degree of mechanistic or structural knowledge can be better addressed with more rationale approaches. The use of target-focussed compound collections (enriched with typical chemical scaffolds of the corresponding target class) has been spear-headed for the kinase target family. For proteases, the use of structure-based drug design had been adopted to guide the H2L process in a more informed manner than by simple SAR exploration around HTS hits. The lack of productivity of HTS for certain target families and the early successes with structure-based drug design have paved the way for fragment based drug discovery (FBDD), which arguably started with the development of the SAR-by-NMR™ technology by the Abbott group in 1996.

Differing concepts for HTS and FBDD

The key conceptual difference between FBDD and HTS driven lead discovery is that FBDD does not rely on potency as the key and sometimes sole driving force for hit compound selection and progression. FBDD rather focuses on the exact fit of a hit compound to a binding site on the protein target. To identify a compound that achieves the most favourable fit, it is argued that the preferred strategy involves the screening of small chemical fragments of compounds that do not come with superfluous functionality to sterically impair the optimal fit with the target. If many of the heavy atoms (i.e. non-hydrogen atoms) of a fragment are directly engaged in the interaction with the target protein, then the fragment is said to be an efficient ligand or binder. Such a small and efficient ligand is a superb starting point for medicinal chemistry, irrespective of its absolute potency. Indeed fragment K is or K ds are frequently in the 0.1 to 1 mM range, which is 2 to 3 orders of magnitude less potent than typical HTS hits. Furthermore, FBDD usually involves a strong element of structural characterisation of the target :: ligand interaction, typically by X-ray crystallography or alternatively by solution based NMR experiments. The structure determination of the target :: ligand interaction is followed by an extensive phase of analysis by molecular modelling in order to derive design ideas for subsequent compound optimisation.

Another important differentiator of HTS and FBDD is the extent of sampling of the respective chemical diversity space which is less comprehensively achieved by lead-like or drug-like compound collections compared to that achieved by collections of fragment-sized compounds. 10 to twenty thousand fragments can readily offer a substantial sampling of the chemical diversity space of fragment sized compounds while, in comparison, even the largest HTS decks of greater than a million compounds only sparsely samples the chemical diversity space for lead- and drug-like chemical matter. However, a key advantage of HTS-based drug discovery is that the greater potency of the hit molecules ensures that the discovery of inhibitors / modulators with cellular activity can be achieved with a reasonable probability of success in a rather short period of time.

Technology selection: HTS or FBDD?

Clearly then FBDD and HTS have different key strengths and we consider that these techniques are complementary rather than (mutually exclusive) alternatives. An organisation may decide to run a HTS campaign against a novel target to rapidly gain access to chemical matter for target modulation and validation in cells or even in vivo. Likewise, the same organisation may decide to run a FBDD project on a well exemplified and highly competitive target where the novelty of the chemical matter and predictability of success in generating lead series compounds are key prerogatives. Additionally, aside from these project specific tactical considerations, there may be technical aspects that predetermine the choice of approach: is there an assay available that suits the requirements for HTS? Is the target amenable to structural biology and has it been previously crystallised? Is the crystal structure of sufficient resolution to support structure-based drug design? Have high-capacity protein production routes been described that give access to sufficient starting material for crystallography and/or NMR?

Areas of overlap between HTS and FBDD

FBDD has fallen within the domain of direct biophysical screening techniques that have a high degree of sensitivity for detecting weak binders with a correspondingly low rate of false-positive hits. Direct biophysical screening techniques are information rich and allow the target :: fragment interaction to be characterised in terms of binding affinity, binding stoichiometry and also binding site if HT crystallography or protein observed NMR are used.

In concert with NMR, we are using a confocal fluorescence correlation spectroscopy approach for fragment screening. This technique was initially conceived for its application to HTS but since then has been adapted for effective fragment screening. This approach combines high-throughput with high-sensitivity and we are in a position to screen large compound files of over 1 million compounds and complement that with fragment screening at an elevated compound concentration. Certain screening groups within the pharmaceutical industry have started to either flag the fragment-like compounds in their HTS screening decks for separate screening at higher concentration or to develop specific fragment screening sets (typically 5,000 to 10,000s). These are routinely screened as part of HTS campaigns (in parallel to the main HTS screening collection), albeit at concentrations in the 0.1 to 0.5 mM range. Eli Lilly and Co, Novartis, GSK, AstraZeneca, Roche, and the Abbott Laboratories amongst others are either applying or seriously exploring the application of high concentration screening of fragment-like molecules by bioassays as part of their screening paradigm. The rationale behind this is to significantly increase the probability of success in the early hit discovery and subsequent H2L optimisation. Often, high concentration screening by bioassay may be accompanied in a complementary manner by well established approaches to FBDD undertaken by specialised biophysics and structural biology groups outside the core HTS screening units.

The benefits of a dual HTS and FBDD approach

We consider that there are at least 3 benefits in pursuing a dual HTS and FBDD approach and these are discussed in the following.

Firstly, there are target-specific, empirical factors and every organisation will have multiple examples for targets that have failed a traditional HTS approach. There are cases for instance in the area of aspartyl and cysteine proteases and in the protein :: protein interaction field that have consistently failed to generate viable leads from HTS. Frequently this can be attributed to a general mismatch of lead-like or drug-like chemical matter and the relevant catalytic pockets (proteases) or interaction hot spots (protein :: protein interaction). In fact, for proteases, the side chain pockets for the substrate specificity residues may be better addressed individually with fragment-sized compounds, with lead compounds being derived by subsequent fragment linking or merging. This is conceptually a better strategy than gambling on a HTS generated hit series that would satisfy the binding requirements of multiple sub-pockets at the same time. There are other targets outside the protease field that show a very good fit to fragment sized molecules and have consequently benefited enormously from the application of FBDD: for example the iNOS, nNOS and eNOS family of enzymes, carbonic anhydrase, D-amino acid oxidase, certain kinases and ATPases (including JNK1, CDK2, Hsp90 and DNA gyrase), hematopoietic prostaglandin D2 synthase, some phosphodiesterases and many more. Not knowing the drugability of one’s target at the outset, a dual-strategy can significantly de-risk the early discovery of chemical starting points without substantially affecting timelines and budget.

Secondly, fragments may represent minimum pharmacophores or discrete pharmacophoric elements and thereby enable various strategies for optimisation for enhanced binding including:

• Computational search approaches (such as in silico docking of available compounds structurally related to the fragment hits)
• Merging and/or linking of relevant fragments
• Fragments growth

Fragment hit data can be used in combination with HTS derived hit compounds to develop preliminary SAR that helps to prioritise the fragment / hit series over e.g. singleton hits. The availability of fragments and HTS derived hits from one chemical scaffold may be considered as an ideal but certainly not unrealistic situation where the transition from screening into structural biology and chemical optimisation can be particularly smooth. From a more technical perspective, the prioritisation of hit compounds for ligand structure determination by crystallography can benefit from a dual fragment / HTS approach, too. In our experience not every target and crystallographic system will lend itself to the determination of ligand structures of weak (i.e., fragment) binders, for instance if the pH and ionic strength for crystal growth or stability are different from the assay used to determine the target interaction in the first place. In these cases the more highly functionalised and potent HTS hit may form a viable alternative for key structural studies.

Thirdly, a dual HTS and fragment strategy enables staged screening, a relatively new concept which is a resource-friendly combination of diverse (Stage 1) and subsequent diverse plus focussed (Stage 2) screening, using the hit information derived from Stage 1 to augment the outcome of Stage 2 with the corresponding hit analogues. Novartis have coined the term “virtual fragment linking” whereby a collection of fragment molecules is first screened at high concentration and based on the results a Bayesian in silico model is developed that is then used to cherry pick thousands of compounds predicted to be active against the target from their HTS screening collection of millions of compounds. This is reported to enable successful focussed screening at lower cost (fewer wells are screened) and higher hit rates.

We advocate the use of a fragment collection for an initial diverse screen to direct the selection of commercial analogues containing substructure features of the fragment hits for supplementation of the HTS compound file used in subsequent HTS. A benefit of this approach is a significant gain in SAR information within the timelines of a HTS project. This benefit arises from the fact that the hit expansion work can be efficiently dealt with on the screening robots, thereby saving resource and time. Additionally, there is a greater level of data consistency achieved if all IC 50s / K is are derived from one seamless process. Researchers at Vertex, pioneers of fragment and structure based drug design, reported on their in-house approach of pre-screening novel targets against fragments, using NMR. No understanding is required of target function, using a NMR binding assay protocol, and the question of reagent development and assay development is only addressed if the NMR screen indicates a high degree of drugability of the target. Thus, NMR is used as a pre-screen to assess the drugability prior to committing resource to a HTS project. Once again, HTS can be augmented with compounds that contain substructural feature of the fragment hits.

Concluding remarks

In conclusion, there are great opportunities for using FBDD, particularly high concentration bioassays, in conjunction with HTS not only for target families that are amenable to structure-based design approaches but also for membrane bound targets where screening of fragments enables rapid generation of SAR information and hypotheses for more efficient H2L. Given the power of this approach and the potential to shorten the early phase of the drug discovery process providing robust lead compound series we expect that this will become a standard approach to hit discovery within the pharmaceutical and biotech industries.