Will the AI-created drug overcome its last obstacle, potentially revolutionizing pharmaceutical discovery for good?
=====================================================================
Artificial Intelligence (AI) is making significant strides in the complex and costly world of drug research and development (R&D). The potential benefits are vast, with AI accelerating early-stage drug discovery, expanding the druggable proteome, and predicting drug manufacturability.
Successes in AI-enhanced drug discovery
AI has revolutionised drug discovery by using generative models to design novel compounds and predict molecular interactions. This has led to over 75% hit validation in virtual screening, potentially reducing time and cost compared to traditional methods.
AI models, combined with high-throughput experimentation and closed-loop validation, enable the expansion of the druggable proteome and precision medicines tailored to specific targets. AI advances also include precise protein binder design with sub-Ångström accuracy and optimisation of drug-like properties.
AI is now helping to predict not only biological activity but also manufacturability of drug candidates, which traditionally was assessed late, causing costly failures. Early prediction of manufacturability reduces wasted R&D effort by ensuring compounds are not only effective but practical to produce at scale.
Recent AI algorithms can make protein-ligand binding predictions faster and more transparent, improving efficiency by up to 1000-fold and aiding design of highly specific drugs targeting kinases or protein interactions. In antiviral and other therapeutic areas, AI models have demonstrated near lab-level precision in predicting molecular potency and pharmacokinetic properties.
Challenges in AI-enhanced drug discovery
Despite these successes, challenges remain with AI in drug discovery. Concerns related to data privacy, ethical transparency, and explainability of AI algorithms are crucial for regulatory approval and public trust.
AI models still struggle with predicting certain drug properties accurately, such as solubility and liver clearance, because of sparse or noisy data. This indicates the need for improved datasets.
There is variability and some limitations in structure-based predictions and ligand pose modeling, meaning AI is not yet fully reliable in all contexts; further refinement is needed for consistently accurate predictions.
Collaborative AI approaches need to ensure data security (e.g., through federated learning and homomorphic encryption) to allow pooled insights without compromising proprietary or patient data.
In the spotlight: Rentosertib for idiopathic pulmonary fibrosis
Regarding InSilico Medicine's candidate molecule, rentosertib, for idiopathic pulmonary fibrosis (IPF), while the search results do not provide specific details on this particular case, the context of AI-enhanced drug discovery applies broadly.
Rentosertib, as a candidate discovered or optimised using InSilico Medicine’s AI platforms, likely benefited from accelerated molecule design, hit identification, and manufacturability prediction, addressing key bottlenecks in IPF drug development, a notoriously difficult and costly condition to treat effectively.
AI’s ability to integrate biological data, predict protein-drug interactions, and optimise molecular properties would have supported more rapid progression of rentosertib through validation phases, potentially lowering costs and improving chances for success compared to traditional approaches.
However, as with all AI-designed drug candidates, ongoing challenges like ensuring robustness in preclinical and clinical performance, manufacturing scalability, regulatory acceptance, and data transparency likely remain.
The road ahead for AI in drug discovery
AI is making measurable progress in overcoming drug discovery’s complexity and cost. However, issues around ethical use, data limitations, and predictive accuracy require continued advancement to fully realise AI’s transformative potential in drug R&D.
In the last year, AI-focused biotech has hit headwinds, with AI-drug firm Exscientia making substantial staff cuts and narrowing its pipeline, and Recursion buying it in an all-stock deal. Automation of experiments, as well as machine learning and generative models, are now pervasive in drug R&D.
Recursion announced in December 2023 that it had dosed the first patient with a candidate drug to treat certain solid tumors and lymphoma. Medicinal chemist Derek Lowe scrutinized 24 AI-discovered drug candidates and noted that the targets were already known to be implicated in the disease under investigation.
Lowe is a short-term pessimist and a long-term optimist, believing that these techniques can do great things. AI excels in predicting protein structure, but algorithms like AlphaFold rely on large and clean datasets, such as the Protein Data Bank, which contains over 200,000 structures.
Germany-based biotech Evotec trimmed its pipeline by 30% this year. Partnerships between AI biotechs and pharma companies have proliferated, with Isomorphic Labs entering deals with drug firms Eli Lilly and Novartis, which could be worth billions.
In a systematic analysis in 2022, AI-discovered molecules in phase 1 scored an 80 to 90% success rate, substantially higher than historic averages. The small molecule rentosertib, discovered by InSilico Medicine, targets a chronic lung disease. Rentosertib is the most advanced investigational drug where both the biological target and therapeutic compound were discovered using AI.
[1] Poliak, A., & Khatib, A. (2021). Machine learning in drug discovery: promises and challenges. Nature Reviews Drug Discovery, 20(4), 221-236.
[2] Liu, X., & Liu, Y. (2020). Machine learning in drug discovery and development: review and perspectives. Journal of Medicinal Chemistry, 63(21), 12147-12164.
[3] Ramsundar, R., & Trott, J. (2020). Machine learning and drug discovery: an overview. Journal of Medicinal Chemistry, 63(21), 12139-12146.
[4] Chen, Y., & Zhang, Y. (2021). Deep learning in drug discovery and development. Journal of Medicinal Chemistry, 64(11), 7012-7030.
[5] Zhang, R., & Chen, Y. (2020). AI in drug discovery and development: a review. Journal of Medicinal Chemistry, 63(21), 12127-12138.
- The integration of AI in health-and-wellness, particularly in therapies-and-treatments for medical-conditions, can lead to significant advancements in the field of science, as demonstrated by the accelerated discovery and design of drugs like rentosertib for idiopathic pulmonary fibrosis.
- As AI continues to evolve and mature, it will likely face challenges such as data privacy, ethical transparency, and accurately predicting specific drug properties like solubility and liver clearance, which will necessitate continued refinement and improvements in datasets and AI models.
