Molecule AI Newsletter: March 2024
Discover the world of in-silico drug discovery through Molecule AI's monthly newsletter. Join our journey of learning and impactful discoveries for a brighter future.
Analysing Quantum bits of drug discovery!
Looking at the pharmaceutical industry’s focus on molecular formations, it is emerging as a natural candidate for the application of quantum computing (QC), which is the next digital frontier. QC is a valuable new technology, although still in its infancy, that is rapidly advancing and will play a significant role in drug discovery and development in years to come.
With its ability to rapidly solve highly complex problems, QC is overcoming the limitations of classical computers, and is expected to enable significant advances across multiple fields. Many believe that quantum processing has the potential to disrupt drug discovery, providing great value by dramatically reducing the >90% failure rate in preclinical and clinical stages.
Perhaps most importantly, beyond accelerating drug discovery and development, QC has the potential to enable innovative approaches to problem solving, opening up the possibility for developing and manufacturing new medications that were not previously thought possible.
As an example, quantum mechanics–enabled synthetic chemistry gives researchers the tools to preclude potentially inactive compounds and to support the synthesis of more challenging compounds. Or that quantum computing has the potential to transform virtual screening through physically precise modelling of drug-target interactions and efficient screening of massive virtual libraries.
Quantum computing may also be useful in the target identification phase by enabling deeper exploration of complex multifactorial diseases that require the modulation of multiple targets.
Sensing these possibilities early on, global pharmaceutical player Novo Nordisk had announced an investment of $200 million, back in 2022, for the first quantum computer for life sciences research. The Novo Nordisk Foundation Quantum Computing Programme, launched in collaboration with the University of Copenhagen, includes researchers in quantum computing from Denmark, Canada, Netherlands and the US.??
Around the same time, Japan’s Fujitsu and RIKEN Research Institute teamed up to create Japan’s first domestically produced and commercialised quantum computer, which eventually hit the market in Spring 2023. Further, Roche has collaborated with Cambridge Quantum Computing to simulate quantum-level chemical interactions with the goal of discovering new Alzheimer treatments and, eventually, candidates for other diseases.
In this list, India is not far behind. The Ministry of Electronics and Information Technology has launched a 14-year roadmap for developing quantum technologies. The roadmap elaborates on the $735 million National Quantum Mission (NQM) that the government had launched in April 2023.
On the whole, there are currently multiple players across the globe intensely exploring this space for drug discovery and development. Major technology companies such as Alibaba, Amazon, IBM, Google, and Microsoft have already launched commercial quantum computing cloud services. It is therefore not surprising that McKinsey estimates global pharma spending on quantum computing in R&D to be in the billions by 2030.
But of course, for most pharma companies interested in exploring QC, the complexities of the technology necessitate the need to establish partnerships with companies that have specialised expertise in this area. It would be a wise decision for drug developers to work in partnership with pure-play QC firms developing software and hardware solutions and offering associated services and support.
Molecule GEN Highlights
Molecule AI is developing Molecule GEN, a web-based, modular drug-discovery platform. Molecule GEN is envisioned as a go-to platform for the drug discovery community, due to its offering of diverse, user-friendly workflows such as the visualisation and analysis of target proteins, generation of high affinity hits against given targets, assessment of the med-chem and ADMET properties of molecules, lead optimisation, docking and molecular dynamics simulations. In this issue of our newsletter, we present to you Molecule GEN's module for de novo target-aware molecule generation using Generative AI: TAGMol.
领英推荐
TAGMol: State-of-the-art De Novo Generation of Small-Molecule Ligands
De novo generation of ligands that bind with known target proteins has the potential to solve one of the most challenging parts of the drug design pipeline. Traditional drug design heavily depends on screening libraries to select suitable lead candidates. However, the screening paradigm limits the possibilities of drugs, as the libraries contain only a tiny fraction of all potential drug-like molecules (on a scale of 1060). De novo generation represents a different approach to the problem by directly generating the potential drug candidates instead of screening. This leads to efficient sampling of potential drugs from the entire search space. Generative AI plays a significant role in achieving this capability, which is not feasible by other computational techniques.?
However, existing generative AI techniques for de novo generation have yet to live up to their promise due to their generation of unrealistic molecules that fail many of the traditional metrics for successful drugs. Therefore, TAGMol, our proprietary solution for de novo generation, was specifically designed to alleviate these challenges and bring de novo generation to the point of practical applicability. We achieve this by using a novel optimisation technique that generates drugs while simultaneously optimising multiple properties. In particular, we consider? 1. Binding energy of the protein-ligand complex with the Vina Dock docking algorithm, 2. Drug Likeness (QED), and 3. Synthetic Accessibility (SA).?
On generating binders for 100 target proteins (previously not seen by the model) covering a wide variety of diseases, TAGMol achieves strong results across the three metrics, as shown in Table 1.
TAGMol represents the best available model for de novo ligand generation and is presently undergoing peer review for the ICML 2024, a premier event in the field of deep learning.
We look forward to telling you about the other exciting capabilities of Molecule GEN in the subsequent issues of our newsletter.
Molecule AI in News