Figure 2.10. Typical control data from two screens. (a) An acceptable assay was designed showing a good separation between signal and background, 3.5-fold and a Z' of 0.86. In contrast, the assay in b performed poorly, with a signal to background of 3.5 and a Z' of —0.2, and it had to be re-designed.

company publications to provide useful starting points for chemistry. However, if there is ormation on the nature or family of targets it are being worked on, then a number of focused approaches can be taken to discover d molecules.

If the target is a known "druggable" target such as a GPCR, ion channel, or simple enzyme, then one could take a "targeted" or stems-based" approach to lead discovery. By collecting and searching all the avail-e compounds in the company compound collection or external commercial libraries with known activity against these family's of targets, the discovery of active compounds can become more effectively targeted. In some companies, these target class compounds sets can be as small as few hundred compounds or as large as several thousands. At this part of the process, CADD techniques can be used most effectively. Sophisticated computer algorithms matching structural information of a biological target or ligand to search internal and external databases to find structural matches predict active site or receptor binding. Alternately, the technique of "virtual screening" can be used to generate virtual libraries of tens of millions of compounds that can be "docked" into the binding site of a protein target of interest. These virtual compound hits can, if possible, be synthesized and tested in a bioassay to confirm the authenticity of the hit (68-70).

3.3.3 High-Throughput Screening. Of course, when the type of information needed to make smart, focused decisions about lead discovery is not available, most pharmaceutical companies now have at their disposal the powerful technology of HTS.

HTS is the process by which very large numbers of compounds (hundreds of thousands) from a variety of sources such as synthetic compound collections, natural product extracts, and combinatorial chemistry libraries are tested against biological targets. The aim of this exercise is to find compounds active against the target. This will include not only compounds in the screening deck that are expected to be active against the target, but also compounds that would not have been predicted to be active. The underlying premise to high-throughput, random screening is that by sheer scale of numbers, you can bias the serendipitous discovery to occur. If the chances of finding a new chemotype against target X is one in a million. then in the HTS world screening one million compounds would maximize the chance of finding an active hit.

This may seem like a simplistic, anti-intellectual approach to drug discovery, but it has proven to be a very successful lead discovery paradigm for decades. The scale and industrialization cf the operation often masks the incredible innovation and smart thinking that has gone into the HTS process over the last 10 years.

High-throughput screening is positioned to have its maximum impact on the earliest part of the drug discovery process, namely lead discovery. The goal of the lead discovery process is to provide a cohort of chemically tractable molecules with sufficiently interesting properties against nominated biological targets. This would trigger further investment of resources in lead optimization programs such as medicinal chemists, disease biology specialists, etc.

A core underpinning technology for the screening process described above is the capability to acquire, store, and rapidly retrieve large numbers of compounds for testing. Most large pharmaceutical companies now use advanced compound management technologies to achieve this goal (71).

The compound management activity covers the acquisition of compounds and libraries, their registration, formatting, QC checking, storage, and retrieval on demand. A typical compound management process would be intimately involved in all aspects of the drug discovery process, supplying compounds in a variety of formats and amounts to lead optimization programs, HTS teams, and biology research teams.

Most pharmaceutical companies have large collections of compounds derived from past and ongoing medicinal chemistry programs and samples acquired from external vendors. In most large pharmaceutical companies, these compounds are stored in large automated storage systems. These compound management systems allowing the storage, reformatting, tracking, and retrieval of compounds, with onward distribution to HTS laboratories or other research laboratories in the company. The size and diversity of the compounds held within a company compoi^nd store largely reflects the size and diversity cf the medicinal chemistry approaches a company has taken over its' history. Because of the recent spate of mergers, some of these compound collections can be over one million compounds, although the average is more likely to be in the range of 300 —500,000. Ideally, the internally acquired and legacy compounds are analyzed before going into the compound store, to select compounds with drug-like properties and eliminate polymers, reactive intermediates, dyes, etc. This refined collection would then form the basis of a "solid" store where compounds exist as dry powders or films in standardized vials. Usually the traditional medicinal chemistry compounds are supplemented with libraries from commercial sources and academic collaborators. Unless an exclusive deal has been negotiated, the compounds available from commercial and academic sources will be available to anyone who wishes to buy them. A competitive if e n e i.

edge can be gained by developing tools to select the "best" external compounds to complement the internal compound collection and by acquiring diverse chemotypes that are absent from the current collection. Chemoinformatic resources and a compound acquisition budget are therefore vital to the maintenance of a modern compound store.

Natural products have historically been a significant component of a compound deck (72). Typically, a crude solvent extract of a microbial fermentation, a plant, or a marine organism was provided to the compound store ready for screening. If a particular extract was shown to have the desired biological activity, it was refermented or re-extracted, fractionated, and the pure compound isolated (73). This isolation typically took weeks to months. The protracted timelines invariably became the "Achilles Heel" of natural products screening because the medicinal chemistry program had either advanced beyond a point where the natural product could not add significant value or the program had been terminated. In an attempt to improve this temporal dilemma, some companies have resorted to pre-fraction-ated crude natural products extracts (74). This tactic involves fractionating natural product extracts before screening to yield a series of fractions that each contains just a few components. By automating this fractionation process, a more efficient overall process can be developed. This approach dramatically decreases the time to reveal an isolated natural product and de-couples the fractionation from the screening process. From our experience, screening less complex natural product extract mixtures tends to reduce the false positive rate in a high-throughput screen. Alternatively, one could purify or purchase purified natural product compounds directly, and then put them through the biological screens as with traditional medicinal chemistry compounds.

A very common source of new compounds for high-throughput screening comes from the prodigious output of combinatorial chemistry laboratories. Large numbers of compounds can be prepared relatively quickly using automated synthesis in either solution or on solid phase. The real art is to build combinatorial libraries that contain "drug-like" molecules that have biological activity. It is beyond the scope of this chapter to detail the various combinatorial chemistries that generate screening libraries.

Having collected and organized a compound deck, the next challenge is to manage the requests from individual investigators both to deposit new compounds from medicinal chemistry laboratories and to ship compounds for biological experiments in a timely fashion. A few tens or hundreds of compounds from the solid compound store can be routinely weighed out using manual or automated systems and delivered to an investigator. However, the supply of the hundreds of thousands required for a high-throughput screen would be totally impractical using the solid store. To overcome this logistical nightmare many compound stores also contain a liquid store. Here, solubilized compounds are stored as individual tubes and in microtiter plates. The solvent of choice is usually dry dimethyl sulfoxide (DMSO). This is a compromise, because not all compounds will readily dissolve in DMSO, but many biological assays can tolerate relatively high concentrations of this solvent. In addition, not all compounds will be stable in DMSO over protracted periods; therefore, there needs to be an appropriate compound refreshment process. It is important to have a high-throughput analytical chemistry capability to monitor the quality of the compounds. The concentration at which compounds are stored very much depends on both the compound handling process, the screening process, and the solubility of the compounds. For example, most cellular assays will not tolerate more than 1% DMSO without adversely affecting the cell physiology, and therefore, a 1 mM stock solution from the compound store will only give 10 joM final concentration in the assay. By storing compounds at a higher concentration, the risk of compounds coming out of solution increases. As a compromise, some groups will routinely store compounds as a 3 mM stock solution in 100% DMSO.

Having compounds available in both tubes and plates is a distinct advantage for screening. The microtiter plates can be readily organized and provided to a high-throughput screening group who will test for biological ac

Figure 2.11. This figure shows some of the components within the Haystack compound storage and retrieval system: (a) the microtiter plate handling system, (b) the solid storage system and the robot that handles the compound vials, and (c) the tube picking robot placing a solubilized compound back into a tube rack.

Figure 2.11. This figure shows some of the components within the Haystack compound storage and retrieval system: (a) the microtiter plate handling system, (b) the solid storage system and the robot that handles the compound vials, and (c) the tube picking robot placing a solubilized compound back into a tube rack.

tivity. Certain compounds that yield the desired activity are repeat tested to measure potency, selectivity, and in vitro toxicity or safety. Because these compounds are likely to be scattered across hundreds of different microtiter plates, a tube store allows you to prepare a customized set of compounds and supply then for screening. Additionally, focused sets cf compounds based around known chemotypes or chemical series are readily assembled from an automatic tube store for lower-throughput screening. Overall, a modern compound inventory management process is highly integrated, requiring a combination of chemistry, information technology, and production engineering skills. Numerous companies now provide very sophisticated compound storage systems that can manage all of the operations described above. An example used at Bristol-Myers Squibb is the Haystack system built by The Technology Partnership, UK (71) (Fig. 2.11).

This particular fully automated storage and retrieval system can store over 750,000 compounds as dry solids and potentially 15 million compounds in a variety of liquid formats.

There have been attempts to specifically engineer combinatorial synthesis approaches to facilitate screening that is more effective by eliminating the need for indirect compound storage and retrieval systems, such as the Haystack (75). In these examples, the compounds were prepared using solid phase synthesis. A compound was synthesized on Tenta-gel beads using both acid and UV sensitive linkers and encoding tags. The tag encodes the order in which monomers were built onto a scaffold such that the exact monomer sequence could be identified. A mixture of approximately 10-20 beads per microtiter plate well was treated with acid, and the released compound assayed for biological activity. If activity was detected, each of the 10-20 beads were separated into individual microtiter plate wells and exposed to UV. If activity was detected, the tag was decoded, identifying the monomer sequence and hence the structure. Using this format, tens of thousands of compounds could be stored in a relatively small store, i.e., a standard refrigerator.

Compound stores also contain screening decks that are available mainly for the HTS laboratories. These screening decks can represent the full collection of compounds in the store or tactically organized subsets (76, 77). The choice of whether to screen a full com pound deck is dependent on the medicinal chemistry insight or structural knowledge of a particular target as discussed earlier.

From a screening perspective, there are significant advantages to a systems based or target class approach. First, similar assays allow paral-lelization of assay design, the so-called "Plug and Play" approach. Second, the screening data for a family of closely related targets is simultaneously generated for set of compounds.

Full compound deck screening is extremely useful if you wish to find new chemotypes, and there is little information on compounds that effect the target of interest.

34 Hit Identification, Profiling, and Candidate Selection

After ordering a copy of the compound deck in the plate format of choice, the screening process can be carried out very rapidly (a matter of weeks). The screening scientist monitors the automated system continuously for both hardware and assay performance. Of all the different stages in the lead discovery process, the actual screening is now the quickest.

3.4.1 Analyzing Screening Hits. The next step is to analyze the primary screening quality control data (66). During a primary screen, a variety of QC plates are inserted into the run, including blank plates, plates containing just DMSO, monitor hardware performance, pipette error, detector misalignments, etc. QC plates containing biological reagents check for drift in the assay over the course of the screen. Typically, panels of known inhibitors, activators, antagonists, or agonists, depending on the assay, are tested at multiple concentrations. Additionally, controls, both negative and positive, are included on each plate. Realtime data analysis allows the screening scientist to continuously monitor the performance of the screen and the robot.

Having analyzed the QC data and eliminated, or repeat tested, any compound plates that failed, the entire screening run is analyzed. It is usual to perform a high-throughput screen with a single replicate of each compound. The results are statistical in nature and interpreted as population data. In the assay validation section (Section 3.2.6), a statistical parameter Z' was introduced. If an assay had a Z' of 0.7 and a known inhibitor was known to cause 50%inhibition at 10 jum, then in a screen, the percentage of inhibition could vary between 35% and 65% because of population statistics. The initial high-throughput screen is normally viewed as a population frequency histogram and as a scatter plot (see Fig. 2.12, a and b, respectively). Having selected a particular cut-off, any compounds with greater than or equal to this value are retested with replicate determinations. Figure 2.13 (a and b) shows the frequency histogram of the retest values and the associated scatter plot. In this particular example, 73% of the active compounds retested in the second assay. Interestingly, the majority of the weak inhibitors, 20-40% inhibition, were false positive compounds that interfered with the assay readout. Figure 2.14 shows the Z' values for this screen's quality control plates.

The vast majority of compounds from a full deck screen has no effect, and statistical analysis allows one to decide when a compound has had a significant effect, designated as a "hit." A hit that seems to be statistically significant could be explained by a variety of reasons; assay false positives, cytotoxicity effects, etc., as well as true pharmacological response. False positives are compounds under test directly interfere with the detection readout, for example, a fluorescent compound or quencher in a prompt fluorescence assay. False positive results arise from pipetting errors that delivered the incorrect amount of a reagent. A second round of assay(s), to analyze the hits eliminates false positive results. For example, assays that indicate which compounds show the desired selectivity against other biological target or lack of cellular cytotoxicity. A third round of screening generates data from concentration response curves so potency, e.g., K; or IC50, and or efficacy for agonists, directs further medicinal chemistry.

3.4.2 Profiling Hits. Increasingly, major companies are adding to the value of high-throughput screening by immediately profiling screening hits against a battery of selectivity, toxicity, and safety assays (78-80). The idea behind this type of extensive profiling is

% Inhibition



■ r ■ . ■'' Tvi •* • . -I • •■ V ..J-.-?'" -V' -



Was this article helpful?

0 0

Post a comment