Introduction

Last week’s article covered AI’s influence in biotechnology and mostly its current influence in speeding up the drug development process. As technology and AI develop, more stages in the drug development pipeline will become automated with hopes of the entire process being automated and accelerated. The next stage after drug discovery is preclinical research. After scientists or machines conduct basic research, they determine if the drug has the potential to enter the market. Drugs that hold this promise are called lead drug targets and enter stage 2,  preclinical research, with the hope that this drug will become FDA-approved and distributed worldwide. In the context of drug repurposing, preclinical applications are important as most repurposed drugs start with preclinical research as previous data points to these candidates being leads.

Deeper Dive into Preclinical Research

Preclinical research is the second longest stage in the drug development process, taking anywhere from three to six years. The main goal of this stage is to evaluate potential treatments for diseases in animals and cells (in vivo/in vitro) and ensure it is safe enough to test with people-driven clinical trials. Various articles have their respective standards for preclinical research; here we will be discussing an overview and simplified five-step process with the information given. Preclinical work starts with target identification and validation, where scientists identify a biological target associated with the illness or diseases they plan to cure and validate that modifying this target will have therapeutic effects. Second, various tests are run and scientists split drugs into failed compounds and hits, thus calling this stage hit identification. As the name suggests, failed compounds are drugs that fall short in any category and can not continue in the drug development pipeline. Hits are compounds identified during the initial screening phase that show desired biological activity against the target.

After hits are found, scientists begin to optimize the compounds with the hope of turning them into leads. Leads are compounds that have been further refined and optimized from their original status as hits. Leads undergo rigorous testing and optimization to address all issues before advancing to clinical trials. After leads are confirmed with in vitro studies (testing taking place in cell lines within a petri dish or test tube), they move on to in vivo studies (using live animal models with strong biological equivalence to humans). The entire hit-to-lead optimization process is slowly becoming automated with technological advancements occurring in both AI and biotechnology. Next, a toxicology and pharmacology test is performed to conduct both short and long-term effects of the drug on animals hoping to simulate its possible effects on humans as much as possible. Finally, scientists start on an Investigational New Drug (IND) application that contains all the preclinical data and is submitted to the FDA review team. If the FDA approves, the drug advances to the next stage - clinical trials; if not, the drug must go through more preclinical research until the FDA deems it safe enough to advance.

For the drug to pass, the FDA requires a pharmacological profile, identification of the acute toxicity of the drug in at least two species of animals, and a short-term toxicity study ranging from two weeks to three months. Throughout this entire stage, the FDA can check if the laboratory is maintaining Good Laboratory Practice (GLP). GLP has regulations set for study conduct, personnel, facilities, equipment, written protocols, operating procedures, study reports, and a system of quality assurance oversight for each study to help assure the safety of FDA-regulated products.

Relevance of Preclinical Stage in Repurposed Drugs

Repurposed drugs are those that have already undergone or are currently undergoing the drug development process. But how much data is needed for a drug to advance through each stage? In the first stage, drug discovery and development, there must be evidence that the compound has potential therapeutic effects on the targeted illness or disease. Most repurposed drugs bypass this stage because scientists begin with a known compound's potential based on existing correlations. For example, inhaled corticosteroids were identified as potential COVID-19 treatments due to their anti-inflammatory effects on the lungs, aligning with COVID-19's respiratory inflammation. Defining the data requirements to expedite or skip preclinical trials is more complex, as many prerequisites must be met before a drug can advance. Essentially, sufficient existing data must demonstrate the drug’s effectiveness, safety, and consistency in at least two animal species (one rodent and one non-rodent). If this data is available, the drug can quickly pass through preclinical trials. With current trends in biotechnology, companies are working to make preclinical trials fully automated using robotics and artificial intelligence.

AI’s Role in Preclinical Development

Once again using Insilico Medicine as an example, they have managed to push their AI-discovered and developed drug to Phase 2 of clinical trials with minimal human assistance showcasing how far technology has advanced. This means all the identification and optimization were either done or discovered by AI. Insilico used their AI platform Pharm.AI and advanced their drug from the nomination of a preclinical candidate to Phase 1 clinical trials within 30 months in comparison to the average time cost which ranges from four and a half to six and a half years. Today, Insilico and partnered organizations have currently put out some information about the drug making it into Phase II clinical trials with high hopes of success. The majority of the basic research before hit identification and optimization has already been automated or is in the process of being automated with machine learning. AI usage was reported in the development of 164 investigational drugs and one approved drug. The two major types of AI used were machine learning and deep learning. These models were used for 12 purposes, mostly drug molecule discovery, which takes place before hit identification. (Druedahl et al. 2024).

Immunocure, one of the many companies working on developing AI to assist in the drug development pipeline, has focused on revolutionizing the hit-to-lead process of preclinical research by applying generative AI. The major benefits of using AI are that it is more effective, efficient, accurate, and reliable, allowing human capital to work in non-automated areas to work simultaneously with machines. The process shown above shows the implementation of generative AI in the scaffold decoration step, but other companies such as Escientia also use generative AI models with Amazon Web Services (AWS) to securely, quickly, and efficiently find drug candidates to accelerate early drug development at a lower cost. RNA sequencing has begun trials on being automated with biotechnology start-ups such as Biostate AI launching their product (something that will be described more in-depth in the next section). One fascinating topic involves the usage of organoids, lab-grown 3D structures that mimic human organs allow AI models to bypass the complicated ethical limitations of animal models allowing deep learning to train these algorithms.

As explained in last week's article, machine learning includes the usage of neural networks, which allows these programs to learn like humans from their previous experiences. In addition to neural networks, there is ensemble learning where multiple neural networks undergo the same experiment to produce better predictions and pass the knowledge on to the next generation of algorithms. Once this process becomes automated, machine learning algorithms will learn how to identify hits and leads, experiment and test them on organoids, and filter data for scientists. Scientists can then determine if they want to advance or withhold the drug. This field is called organoid intelligence and it is becoming recognized by both government and major companies with a projected growth rate of 22% between 2023 and 2030. (Wei 2023).  AlphaFold 3, the newest model of Google’s AlphaFold, can predict the structure and interaction of all life’s molecules with high levels of accuracy. Google quotes, “at least a 50% improvement compared with existing prediction methods, and for some categories of interaction we have doubled prediction accuracy.” This would speed up the in vivo and in vitro testing and a combination of all these technologies would make a drastic decrease in both time and cost through the drug development process.

Current Media with Preclinical Development

In recent news with AI and preclinical research, the process of RNA sequencing is becoming automated. “AI-based healthcare startup, Biostate AI, announced the launch of two service products total RNA sequencing and OmicsWeb Copilot for analyzing RNA sequencing data.” (ETHealthworld 2024). Like Biostate AI, multiple smaller startups are developing AI models to aid in the drug development pipeline with hopes of decreasing both costs and time throughout the process. Total RNA sequencing helps speed up the process by quantifying which genes are important to the host cell for the scientists to target while making the compound. Biostate AI is one of many companies that have joined in on making machine learning algorithms in different steps of preclinical research. This has also led major companies to outsource the research and development of their novel drug products to contract research organizations (CROs).

Cresilon just compared their first phase of the preclinical study with the help of the US Department of Defense, examining the company's hemostatic hydrogel, which is designed to stem bleeding and reduce damage in the immediate treatment of traumatic brain injury. In this specific case, it was concluded from the successful preclinical phase, that the best method of delivery was a syringe and the plant-based gel mitigated the effects of a traumatic head or brain injury.

Written and Constructed by Joshua Minami, Christopher Korban, Christian Chung

• Ai in drug discovery and development - will it live up to the hype?. AI In Drug Discovery And Development - Will It Live Up To The Hype. (n.d.). https://www.drugdiscoveryonline.com/doc/ai-in-drug-discovery-and-development-will-it-live-up-to-the-hype-0001

• Author links open overlay panelR.S.K. Vijayan a, a, b, Highlights•The use of AI-driven solutions to enable pre-clinical drug discovery is growing steadily within the pharmaceutical industry.•AI technologies can be leveraged across the drug discovery value chain.•Artificial intelligence may transform the way w, & AbstractArtificial intelligence (AI) is becoming an integral part of drug discovery. It has the potential to deliver across the drug discovery and development value chain. (2021, November 25). Enhancing preclinical drug discovery with Artificial Intelligence. Drug Discovery Today. https://www.sciencedirect.com/science/article/pii/S1359644621005043

• Commissioner, O. of the. (n.d.). Step 2: Preclinical research. U.S. Food and Drug Administration. https://www.fda.gov/patients/drug-development-process/step-2-preclinical-research

• Dahlin, J. L., Inglese, J., & Walters, M. A. (2015, April 1). Mitigating risk in academic preclinical drug discovery. Nature News. https://www.nature.com/articles/nrd4578

• Drug repurposing strategies, challenges and successes. Drug Discovery from Technology Networks. (n.d.). https://www.technologynetworks.com/drug-discovery/articles/drug-repurposing-strategies-challenges-and-successes-384263

• Hit to lead identification - immunocure: Ai Powered Drug Discovery Cro - computational chemistry services & Synthesis - CRO services. Immunocure. (2024, January 17). https://immunocure.us/hit-to-lead-identification/

• How organoids can redefine pre-clinical research. Drug Target Review. (2023, September 19). https://www.drugtargetreview.com/article/111597/how-organoids-can-redefine-pre-clinical-research/

• InSilico Medicine. (2023, January 10). Insilico medicine announces positive topline results of the New Zealand Phase 1 trial of ins018_055, an AI-designed drug for an AI-discovered target. GlobeNewswire News Room. https://www.globenewswire.com/news-release/2023/01/10/2586249/31533/en/Insilico-Medicine-announces-positive-topline-results-of-the-New-Zealand-Phase-1-trial-of-INS018_055-an-AI-designed-drug-for-an-AI-discovered-target.html

• Jonker, A. H., O’Connor, D., Cavaller-Bellaubi, M., Fetro, C., Gogou, M., ’T Hoen, P. A. C., de Kort, M., Stone, H., Valentine, N., & Pasmooij, A. M. G. (2024, January 5). Drug repurposing for Rare: Progress and opportunities for the Rare Disease Community. Frontiers. https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2024.1352803/full

• Macdonald, G. J. (2024, January 12). Laboratory automation reaches every stage of drug development. GEN. https://www.genengnews.com/topics/artificial-intelligence/laboratory-automation-reaches-every-stage-of-drug-development/

• Preclinical regulatory requirements. Preclinical Regulatory Requirements | Social Science Research Institute. (n.d.). https://ssri.psu.edu/clinicalresearchguidebook/preclinical-regulatory-requirements#:~:text=At%20the%20preclinical%20stage%2C%20the,two%20species%20of%20animals%2C%20and

• team, G. D. A. (2024, May 8). Alphafold 3 predicts the structure and interactions of all of life’s molecules. Google. https://blog.google/technology/ai/google-deepmind-isomorphic-alphafold-3-ai-model/#life-molecules

• Timonera, K. (2024, June 26). Ai lead scoring: Drive Revenue with intelligent automation. eWEEK. https://www.eweek.com/artificial-intelligence/ai-lead-scoring/

• Weth, F. R., Hoggarth, G. B., Weth, A. F., Paterson, E., White, M. P. J., Tan, S. T., Peng, L., & Gray, C. (2023, November 27). Unlocking hidden potential: Advancements, approaches, and obstacles in repurposing drugs for cancer therapy. Nature News. https://www.nature.com/articles/s41416-023-02502-9

• www.ETHealthworld.com. (2024, July 10). Healthcare Startup, Biostate AI launches free for use RNA sequencing data analysis tools - ET healthworld. ETHealthworld.com. https://health.economictimes.indiatimes.com/news/health-it/healthcare-startup-biostate-ai-launches-free-for-use-rna-sequencing-data-analysis-tools/111627946