Using Cresset to grow and link distant fragment hits with sensible chemistry
The main advantages of the fragment based drug discovery paradigm are (1) the starting points, despite having weak potency, have high efficiency of interaction with biological targets, and (2) optimisation of the fragment by adding atoms is usually done in a way which tries to preserve the initial efficiency, to yield a tractable lead with superior properties.
One disadvantage is that modern hit optimisation strategies, incorporating cheminformatics and molecular modelling techniques, are deficient at a number of levels in the way they handle these quite small molecular fragments. Consequently the use of computational techniques to prioritize new molecules for synthesis is patchy with many organisations developing and using specialist applications to address the perceived weakness in commercial methods.
Our computational approach uses the shape and electrostatic similarity of ligands to relate and interconvert molecules in a way that is relevant to their behaviour in biological systems. We have shown that our molecular description has high utility in virtual screening, ligand alignment and scaffold hopping. We have also reported how our approach is useful for describing the molecular space of commercial fragment libraries and how this can be used to guide the selection of screening libraries.
An extension to our existing scaffold hoping methodology (embodied in the product ’Spark’) where a fragment lead is grown computationally to introduce additional functionality with high probability of success was described in 2012. This modification used the information from other known actives to describe new binding pockets that were accessible to the chosen starting fragment which was then grown to fill these new pockets. Critically the moieties chosen to grow the fragment were selected from databases of available reagents ensuring a high probability of chemical synthesizability as well as activity.
We now describe a further extension to our fragment growing methodology to aid fragment optimisation through linking fragments that have been shown to bind to distinct (distant or adjacent) pockets. The use of databases of moieties derived from real compounds gives a high degree of diversity combined with reasonable probability of chemical synthesis. Further control over the chemistry that will be employed is gained by manual selection of connected atoms that form the new bonds to the existing fragment hits. Creation of final ‘product’ molecules which are fully energy minimized before computational scoring ensures that only linking fragments that can truly work are progressed into the final results. The method will be discussed in detail including the presentation of relevant case studies.
One disadvantage is that modern hit optimisation strategies, incorporating cheminformatics and molecular modelling techniques, are deficient at a number of levels in the way they handle these quite small molecular fragments. Consequently the use of computational techniques to prioritize new molecules for synthesis is patchy with many organisations developing and using specialist applications to address the perceived weakness in commercial methods.
Our computational approach uses the shape and electrostatic similarity of ligands to relate and interconvert molecules in a way that is relevant to their behaviour in biological systems. We have shown that our molecular description has high utility in virtual screening, ligand alignment and scaffold hopping. We have also reported how our approach is useful for describing the molecular space of commercial fragment libraries and how this can be used to guide the selection of screening libraries.
An extension to our existing scaffold hoping methodology (embodied in the product ’Spark’) where a fragment lead is grown computationally to introduce additional functionality with high probability of success was described in 2012. This modification used the information from other known actives to describe new binding pockets that were accessible to the chosen starting fragment which was then grown to fill these new pockets. Critically the moieties chosen to grow the fragment were selected from databases of available reagents ensuring a high probability of chemical synthesizability as well as activity.
We now describe a further extension to our fragment growing methodology to aid fragment optimisation through linking fragments that have been shown to bind to distinct (distant or adjacent) pockets. The use of databases of moieties derived from real compounds gives a high degree of diversity combined with reasonable probability of chemical synthesis. Further control over the chemistry that will be employed is gained by manual selection of connected atoms that form the new bonds to the existing fragment hits. Creation of final ‘product’ molecules which are fully energy minimized before computational scoring ensures that only linking fragments that can truly work are progressed into the final results. The method will be discussed in detail including the presentation of relevant case studies.
