Skip To Content

Top 3 Trends in the World of Physicochemical Properties

October 29, 2024
by Bara Townsend, Marketing Communications Specialist

The Key Takeaway Messages From This Year’s PhysChem Forum

The annual PhysChem Forum meeting brings together scientists from industry and academia to discuss topics around physical chemistry, and its role in directing absorption, distribution, metabolism, excretion (ADME), and toxicity properties of chemical compounds.

The 21st PhysChem Forum was hosted by Syngenta, with a focus on ‘Closing the loop—measuring, modelling, and predicting physicochemical properties’, and we asked Andrius Sazonovas, the Director of Percepta Solutions, to share his thoughts on the current hot topics in the world of physical chemistry and physicochemical properties.

1. Beyond Rule of 5 (bRo5)—Balancing Lipophilicity and Permeability

Lipinski’s ‘Rule of 5’, was originally published to provide scientists with a guideline for identifying orally bioavailable drug candidates.1 These rules changed how the industry approaches the drug design process, forcing research chemists to formally consider the physicochemical properties of compounds, such as lipophilicity and ionization. Whilst Dr. Lipinski’s work was fundamental at the time, he stresses that these rules were merely guidelines designed to streamline lead optimization and speed up the drug discovery process.

As an increasing number of pharmaceutical companies analyzed their own data, it became apparent that there are many orally bioavailable pharmaceutical candidates which don’t comply with the Ro5. This led to scientists moving out of the constraints of Ro5, into the ‘beyond the rule of 5’ (bRo5) chemical space. This trend is likely to continue as pharma/biopharma look to broaden the space of druggable diseases with biologic therapies, accelerate drug discovery and development, and lower R&D costs.2-6

Novel Modalities in the Changing Landscape of Drug Discovery

PhysChem Forum has focused on bRo5 chemical space in previous years, and although it wasn’t the only focal point this year, it was clear that bRo5 compounds are still of high interest. The lecture from the keynote speaker, Dr. Marianne Ashford (AstraZeneca), was a testament to this as she explained how the development of novel drug modalities is changing the face of drug development, from discovery through to delivery.

Dr. Ashford highlighted how the drug development landscape has gradually changed, from designing therapeutics for symptom management to now focusing on stopping, slowing, and eliminating diseases, such as cancers. Marianne also shared insight into how the focus has shifted from targeting proteins to instead targeting DNA/RNA sites with nucleotide-based therapeutics, such as lipid-nanoparticles (LNPs). She also emphasized that drug delivery methods are also being reviewed, with these now moving towards site specific, intracellular delivery. However, even the most cutting-edge research into drug delivery routes is not capable of providing a silver bullet solution for the pharmaceuticals with flawed fundamental properties. Hence, the need to thoroughly evaluate a compound’s full physicochemical profile as early as possible in the discovery stage is not going away anytime soon.

Marianne also discussed the diversity of the drug modalities which are going to shape the next generation of therapeutics. These new modalities are likely to range from advanced small molecules, such as PROTACs and antibody-drug conjugates, to peptide-based conjugates and cell therapy.

The key take-home message from Dr. Ashford’s presentation was that although more complex medicines will enable expansion of the druggable target space, thus treating more patients, more measurements and predictive modelling in the early stages of development will be required to ensure efficacy and efficient intracellular delivery.

Utilizing reliable property prediction software is a great way for scientists to consider PhysChem profiles in early discovery, offering un-biased structure design and lead optimization that is fully based on ideal property profiles of the target compound.

2. The Argument Against Relying on AI and ML for Complex Predictions

With the recent advancements in artificial intelligence and machine learning, there is an increasing interest in utilizing these tools for PhysChem predictions. However, the presentations at this year’s PhysChem Forum served as an important reminder to sometimes take a step back from the complex, sometimes overcomplicated, models and go back to using fundamental classical calculations.

Dr. David Palmer’s presentation clearly supported this approach as he discussed his group’s approach to developing solvation-based molecular descriptors for predicting physicochemical properties.7 His research largely focuses on predicting the solvation behavior of compounds using a statistical mechanics theory-based Reference Interaction Site Model (RISM). Dr. Palmer highlighted how simplifying their models to perform calculations based on long forgotten one-dimensional, rather than 3D, approximations in combination with deep learning technologies still achieved the sufficient accuracy of their predictions without the compromise of extensive additional computational expenditure.

Dr. Palmer’s talk was a stark reminder of the concepts that can also be found formulated in the already existing Good Machine Learning/Artificial Intelligence Practices (GMLP/GAIP)—namely, do not unnecessarily overcomplicate things without understanding how that will affect the accuracy and applicability/interpretability balance. So, while the “traditional” methods, such as structural fragmentation, PLS regression, physicochemically based mechanistic modelling and others, still offer solid foundations for predictive approaches in the field of PhysChem, ADME and Tox properties, a more intuitive and viable strategy would be to incorporate these to work with any novel AI and ML technologies rather than being replaced by them.

Our Percepta software offers the best of both worlds—traditional methods for property prediction combined with ML tools, allowing drug discovery teams to expand their chemical space coverage to include novel chemistry by training the algorithms with their own data, eliminating the need to build models from scratch.

3. PhysChem for More Than Just Pharmaceuticals

As Dr. Chris Baker and Dr. Jonathan Rains from Syngenta shared, agrochemicals have a wide range of applications, such as herbicides, pesticides, and fungicides, all of which require careful design from the start of the development process. Although the requirements for lipophilicity, solubility, or ionization of agrochemicals may be different to pharmaceuticals, the research and development processes share many of the same challenges. The agrochemical researchers must consider the physicochemical and ADMET properties during lead optimization to identify promising candidates and ensure these have the correct properties and efficacy for their intended purpose, whilst also remaining safe for plants, animals, humans, and the environment.

Dr. Baker talked about Syngenta’s drive to go digital which led to the development of their own predictive algorithms. Syngenta are able to constantly train these with their own periodically emerging new experimental data, meaning they are able to confidently classify compounds within the chemical space of their interest even prior to synthesis, highlighting the need for reliable PhysChem predictors to speed up their lead compound discovery outside the pharmaceutical sector.

Syngenta welcoming the PhysChem Forum to their main UK R&D site served as a reminder that industries other than pharmaceuticals often focus on physicochemical properties during product development. In the world of food and beverages, physchem properties may impact the stability, shelf life, and even safety of foods. In the fragrances and formulations sector, ability to predict properties, such as the partition coefficient (logP) or toxicity, helps evaluate the fragrance deposition, predict safety in case of ingestion, whilst also reducing the need for animal testing. These are just a couple examples that emphasize why it is important to consider physicochemical properties in industries other therapeutic drug development.

Software for Physicochemical Property Prediction

ACD/Labs offers predictive software to empower scientists with information for data-driven decision-making without the need for measurement of physicochemical properties. PhysChem Suite provides tools to help you investigate structure modifications that meet the target property profile in addition to calculators for lipophilicity (logP, logD), ionization (pKa), aqueous solubility, rule of 5 compliance, and more.

References

  1. A. Lipinski, F. Lombardo, B. W. Dominy, P. J. Feeney. (1997). Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev, 23, 3-25. DOI: /10.1016/S0169-409X(96)00423-1
  2. M-Q. Zhang, B. Wilkinson. (2007). Drug discovery beyond the ‘rule-of-five’. Opin, Biotechnol, 18(6), 478-488. DOI: 10.1016/j.copbio.2007.10.005
  3. DeGoey, H-J. Chen, P. Cox, M. D. Wendt. (2018). Beyond the Rule of 5: Lessons Learned from AbbVie’s Drugs and Compound Collection. J Med. Chem., 61(7), 2636-2651. DOI: 10.1021/acs.jmedchem.7b00717
  4. J. Young. (2023). Today’s drug discovery and the shadow of the rule of 5. Expert Opin Drug Discov., 18(9), 965-972. DOI: 10.1080/17460441.2023.2228199
  5. V. Hartung, B. R. Huck, A. Crespo. (2023). Rules were made to be broken. Nat. Rev. Chem., 7, 3-4. DOI: 10.1038/s41570-022-00451-0
  6. B. Halford. (2023). Wrestling with Lipinski’s rule of 5. C&EN, 101(8). Read article
  7. D. J. Fowles and D. S. Palmer (2023). Solvation entropy, enthalpy and free energy prediction using a multi-task deep learning functional in 1D-RISM. Phys. Chem. Chem. Phys., 25, 6944-6954. DOI: /10.1039/D3CP00199G

Send me more info!

Subscribe to receive more information about Property Prediction from ACD/Labs

This field is for validation purposes and should be left unchanged.

Comments

Your email address will not be published. Required fields are marked *

Send me more info!

Subscribe to receive more information about Property Prediction from ACD/Labs

This field is for validation purposes and should be left unchanged.