Fragment screening (FS) become nowadays rather a well-accepted source of hits in the process of drug discovery. It is now a common practice, particularly within large pharmaceutical companies to run FS in parallel with HTS (high-throughput screening). It has a number of potential advantages over screening larger compounds as an approach to the identification of starting points for drug development. Although, in contrast to larger compounds, a fragment typically has fewer interactions with the target protein and thus lower affinity overall, thermodynamics suggest that these fragment-protein interactions are individually of greater energetic reward. Additionally, screening can more effectively sample the chemical space of smaller, less complex compounds, which improves the odds of identifying binders that have high ligand efficiency. Fragments may thus provide medicinal chemists with a greater number of promising opportunities and, owing to their smaller size, greater flexibility in the optimization process.

Generally, fragment-based (FB) approaches present a multistep workflow (see Fig.1) involving different scientific backgrounds and a wide range of biophysical, biochemical and biological techniques.

Figure 1. General workflow of fragment-based drug discovery divided into four phases with indicated disciplines that play major roles in the realization of each phase.

The first step of this approach involves the selection of the fragment set. In our opinion, fragments should be kept relatively small in order to maximize their potential interactions with proteins. The collection components are normally chosen considering the Congreve’s Rule of Three. This “Rule” has always been interpreted more as a guideline resuming the optimal physicochemical properties a molecule should possess to be considered a suitable fragment. These physicochemical properties can be listed as follows: Molecular Weight ≤ 300 Da, clogP ≤ 3 as well as the number of rotatable bonds and hydrogen-bond acceptors (HBA) and donors (HBD), Polar Surface Area ≤ 60 Å2. One of the most important properties required is high solubility of the compounds. It needs to be taken into account that the compounds will be screened in µM or even mM concentration. Since the screening methods are not very robust in FS, the compounds are combined in so-called cocktails consisting of 3-10 diverse structurally compounds (see Fig. 2).

Figure 2. Cocktails of fragments are often used to increase screening throughput.

I would like to spend a few more sentences focusing on the initial selection of the fragments for the screening deck. The fragment screening library is accessible in almost any vendors portfolio. In most cases, it is selected in a very simple way by applying the rule of three filters on the already available building blocks. Some of the vendors pay a bit more attention to the selection of the fragment libraries and group them according to their diversity, 3D-like character or dedication to screening technology. At Selvita we are of the opinion that the screening deck of fragments can be kept modest but chosen with great caution. Our chemoinformatic platform allowed us to prepare a set over 3000 fragments that has a balanced distribution of nitrogen atoms, oxygen atoms presented in various functional groups. It is important in the context of a recently published review that reports observed topological and functional-group diversity of fragments coupled with their polarity-lipophilicity balance in fragment-protein complexes deposited in PDB. We have noticed the fragments that are discussed in this review are biased towards the “old fashioned” paradigm of flat and heteroaromatic moieties as the only right choice of structures suitable for medicinal chemists in drug discovery. It is probably due to the retrospective analysis of historical data. The screening collection paradigm has evolved in recent years as well for screening compounds, not just for screening fragments. Unfortunately, for the proof of concept of usefulness of more 3D-like character of fragments reach in sp3 hybridization atoms we would still need to wait for a couple of years. Nevertheless, Selvita suggests to pay attention to the quality (that includes diversity and design criteria) of the screening deck. The screening fragments could be considered as simple building blocks but they can also be employed as scaffolds that are further diversified on one of the available diversification points to yield an additional set of fragments suitable for screening as illustrated in Figure 3.

Figure 3. Example of building blocks and their diversification to create a proprietary fragment collection. Each central spiro building block is diversified on only one possible exit point using various type of reactions resulting in a very small subset of fragments that follows medicinal chemistry philosophy.

If we invest in designing a novel scaffold with two or more exit vectors, using simple parallel chemistry we diversify the scaffolds on the orthogonal position, we can have a very proprietary collection. In return, the optimization process of phase 2 and 3 will be more rapid since the chemistry will be tamed and will give the medicinal chemist an ample of opportunities to work with the hits and later on leads. Another advantage of this approach is the IP situation for novel scaffolds.

The second step of the workflow consists of the evaluation of a primary screening of the collection against the chosen target. The employment of a suitable method is dictated by the target and the possibility of detection of weak binding events by sensitive biophysical techniques. Such techniques are for example X-Ray Crystallography, 1D-/2D-NMR, Surface Plasmon Resonance (SPR), Fluorescence Polarization (FP), IsoThermal Calorimetry (ITC), MicroScale Thermophoresis (MST) or Differential Scanning Fluorimetry (DSF). At least one of the methods should be able to indicate the binding place of the hit fragment. In general, FS delivers higher hit rates than HTS, mainly due to the fact that large compounds have a greater chance of a mismatch with the target protein binding site. Additionally, since the size of chemical space increases exponentially with molecular size, statistically, covering chemical space is more manageable with fragments than with drug-sized molecules. High concentration of screen fragments may also lead to false positive hits. Therefore, hits need to be confirmed using multiple biophysical techniques, highlighting the benefit of using orthogonal detection methods for confirmation of fragment binding. Here the well thought through triaging process is not to be underestimated. Selvita’s successful FBDD story consists of implementation of MST to perform primary screening, followed by SPR and later X-ray as orthogonal, confirmatory screening techniques was heavily supported by chemoinformatic tools. This approach allowed to nominate validated fragment hits ready for further optimization and transformation to lead-like sized molecules.
Mentioning that we are making a step into the next phase indicated in our FBDD workflow, namely the hit evolution. The objective of this phase is to design potent inhibitors by either growing the fragment, linking fragments, or merging fragments onto existing scaffolds for the target. In practice, all of the mentioned strategies are exploited with great help from computation chemistry tools. We experience that a tight collaboration between medicinal chemists and computational chemists involved in FBDD project is a key to success. Noteworthy, in silico techniques are indeed establishing more and more their position in FBDD programs in reason of their rapidly raised efficiency that is translated in very rapid execution and high hit rates. Faster, more powerful computers and more accurate software shall not be neglected, especially when operated by a knowledgeable computational chemist with experience gained during versatile involvement in various drug discovery programs.
The fourth phase that is Lead Optimization does not differ from any other campaign of drug discovery regardless of a source of initial hits.

At this stage, we would like to share with you with our observation. The origin of fragment-based drug discovery lays in the way how medicinal chemists used to treat their full-sized molecular hits. Most of the experienced medicinal chemists would start their work by truncation the hit in order to define the smallest and still active fragment. Once this was established the medicinal chemists in charge would start to diversify this fragment by adding other important moieties that would yield a high-quality chemical matter obeying rules defined by target product profile.

Anna Karawajczyk, Ph.D.
Team Leader, Computational Chemistry

To contact the author please email at