I am reading sporadically through In Silico Drug Discovery and Design: Theory, Methods, Challenges and Applications edited by Claudio N. Cavasotto, PhD. I will be sharing my notes and some thoughts and ideas as they come up, which will emerge out of order (temporally) as I am looking out for some things specifically. Everything here is my paraphrasing from the book.

Section II – Advanced Techniques

Chapter 14: Fragment-Based Methods in Drug Design

• High-throughput screening is great but not a panacea
• Fragment-based drug design offers several advantages
  • A high hit rate
  • Allowing for lead generation with more optimal properties
• Different methods
  • Biophysical – NMR, X-ray crystallography, SPR (SPR can process the greatest number of fragments)
  • Biochemical, which can process more fragments than NMR or X-ray crystallography

Section III – Challenges

Chapter 17: In Silico Approaches Assisting the Rational Design of Low Molecular Weight Protein-Protein Interaction Modulators

• We have generally been targeting protein-protein interactions (PPIs) using biologics and other peptides. This works but small molecules still have their advantages.
• There are many challenges in the drug development process as-is, and they persist (and get amplified) when looking for modulators of PPIs.
  • Choosing the right target is essential, and not made any easier by there being even more possible PPIs than proteins (trivially)!
  • To make matters worse, the interfaces where PPIs occur are physically much harder to drug than the active sites of enzmyes or receptors which we’re used to finding ligands for
  • Traditional compound libraries aren’t necessarily suited for turning up any PPI modulators. There’s probably some gains to be had just by curating libraries for this purpose.
  • Good knowledge of the protein’s 3d structure, especially at the interface, is essential.
  • Modulators of PPIs, when they are found, tend to have unfavourable properties such as being big and very lipophilic, which increases the chance for toxicity.
• A (very clever, in my opinion) way to find targets is to look at SNPs that affect a given PPI.
• The traditional delineation between the two main ways of molecular screening are relevant here as well
  • Ligand-based screening starts with molecules with known interactions and then finds other molecules likely to act similarly
  • Structure-based screening has the computer simulate docking of the compound library items to the protein of interest and picks the best ones
• There’s hope for modulating PPIs based on some drugs already discovered
  • Molecules targeting the VEGF-VEGFR interaction, the EF-calmodulin interaction and the IFN-alpha action on its receptor have been found, assisted by in silico methods

Chapter 18: Incorporating Binding Kinetics in Drug Design

• Binding kinetics can make a big difference sometimes: features such as slow dissociation are sometimes required for sufficient efficacy
  • On the flip side, sometimes a faster dissociation is superior. This is often desired (or simply found beneficial) in CNS drugs, presumably (ie this is me making it up) because of the sensitivity to moment-to-moment changes in endogenous agonist levels that is inherent to much of the system. The authors here use memantine as an example for fast-off kinetics at NDMAR, and then go on to mention atypical antipsychotics which are fast-off at D2R. They don’t mention clozapine themselves but the paper they cite does. I remember reading about its binding kinetics as being the reason for why it (and other atypicals) are less likely to cause extrapyramidal sides and being very impressed. Of course clozapine wasn’t designed using the modern methods described here, but alas.
• Up to the writing of the book there hasn’t been a huge amount done about any of this, in fact there isn’t much in the way of guiding principles that can be relied on to design molecules with specifically desired kinetics
• That said, we can still do better than guessing. The molecule can be modified to stabilize the state where it is bound to the target more than the transition state (the state in the process binding). Once the molecule is bound, the path to dissociation will be more energetically disfavoured.
• There are experimental methods that can be used to investigate binding kinetics but the process can be much lower cost overall if computation is used upfront to reduce the amount of physical work needed.
• Section 18.4 describes the so-called “mining-minima approach” where a simulated annealing (heating and cooling) process of the ligand bound to the protein is used which effectively ends up sampling many possible configurations. The heating lets the simulation ‘find’ configurations that could be very different (‘far away’) from the starting configuration.
  • These configurations can be put together to figure out different docking pathways.
• Section 18.5 describes some math for one method that isn’t completely beyond me but still too complex for me to summarize