This blog post is jointly written by Lucian, Joe and Susan who recently attended the Spring ACS National Meeting 2019.
The Spring ACS National Meeting was held in sunny Orlando, Florida and was a five day event (29th March – 4th April). The temperature averaged 25°C , which was amazing compared to the UK (sorry) and meant we all got a lovely tan. We all presented our work in the form of talks in the divisions of COMP or CINF but in this blog post we write about our highlights of the conference.
Lucian
Title: Chemibodies — Future Therapeutics?
Background:
Small molecules are traditionally used in targeting deep pockets in the protein, while antibodies are highly effective at binding protein surfaces. They have different advantages and limitations as therapeutics. For instances, the targets are only partially inhibited by the antibody in some cases, and the small molecule have limitations in terms of efficacy and side effects. To overcome these limitations, Cheng et al. introduce “chemibodies”, which leverage properties of both small molecule and antibody to target the same protein.
Small molecule + Antibody = Chemibodies
General Idea:
A small-molecule-antibody hybrid is used to target the same biological target at dual binding sites. This approach preserves advantages of antibody (longer half-life and high selectivity) and the small molecules can access active sites that are NOT accessible to antibodies.
Procedures:
Three major steps are required to build chemibodies:
- Identification of conjugation site
- Linker design
- Small molecule design
The detail of the computational analysis can be found in (1).
DPP-IV enzyme was used as a proof of concept study. They showed superior specificity, potency and pharmacokinetic properties than either antibody or small molecule alone.
I believe there is potential for other targets such as ion-channels and transporters and this approach should be further investigated.
Reference
(1) Cheng et al. Structure-guided Discovery of Dual-recognition Chemibodies Scientific Reports volume 8, 7570 (2018)
Joe
One talk that stood out for me looked at using molecular dynamics simulations to locate cryptic binding sites, from the lab of Francesco L . Gervasio :
The talk stemmed from their recent publication which initially found that extended MD simulations, and even solute tempering, failed to pry open known cryptic binding sites on apo structures. They also noted that when you initiated an apo simulation from its ligand bound conformation, the pocket was quickly lost. It was then observed that if they performed the simulations in a benzene solution, some cases demonstrated an increase in cryptic binding site exposure.
Their solution started to take shape with the development of their novel “SWISH” approach (Sampling Water Interfaces through Scaled Hamiltonians). Given that most cryptic binding sites are apolar and hydrophobic, they sought a method to replicate their previous observations of small molecules prying open the cryptic pocket. This was done by making alchemical alterations to the protein-water interactions. Through scaling the non-bonded interactions of the solvent and protein, the interactions of the water with the protein become more protein-ligand like. Combining this technique with solutions containing small molecular probes, they found an increase in overall pocket exposure and a novel way of exposing potential cryptic binding sites.
I am particularly excited by the future of this approach, especially as data presented showed that through altering the small molecules in the solutions to be substructures of known binders, previously unknown binding sites of ligands could be identified.On a different tangent, I would recommend also looking at the work done by Chaya D. Stern et al. They have been investigating how Wiberg Bond orders could be used to reduce the cost of QM torsion scanning. Her slides from ACS can be found online here.
Susan
The conference was large and in tandem to the various parallel talks given by each division, there was opportunity to attend career workshops which were very useful and quite popular with the student attendees. One of the workshops which I went to was one on how to network, where they gave us examples of what to say and we could practice on each other.
Some of my highlights from the talks I attended include “Computational Case Studies” given by David Koes (University of Pittsburgh). Indeed, like his abstract suggested, this talk was not one to attend if you wanted a happy ending. He talked us through his workflow (MD, pocket analysis, fragment docking, pharmacophore query, selecting and ordering candidates from a chemical database then experimental screening) and how it fared in four case studies. In all, the experiments gave rise to “hit” molecules which showed activity against the target of interest but for many, further SAR did not fully support the original predicted binding modes.
Connor Coley (MIT) gave a well attended talk where he presented his work on predicting chemical reactivity (based on their paper). His method uses a graph-convolutional neural network however, the model offers interpretability due to the global attention mechanism, as you can see where the model predicts sites of reactivity, much like how human chemists would rationalise reactivity.
I was intrigued by the title of a talk given by Terry Stouch (Science for Solutions), “Mindfulness and Care of the Foundation of AI”. Mindfulness – or being aware of yourself and your surroundings – makes sense in terms of on your Machine Learning / AI data. Poor data = poor model = mislead decisions. He talked about being aware of potential data creep and getting metadata if you can.
