The Drug Discovery Bottleneck: Why We Need AI
Drug discovery is a long and arduous process that is typically likened to a marathon, rather than a sprint. Initially, a scientist will identify a disease target and conduct a large-scale screen of chemicals to identify the most promising compounds that interact optimally with the target. Once a number of leads have been identified, the compounds are then optimized to possess the best properties for a potential drug, which is then put through a number of preclinical disease models in order to test whether it has the desired pharmacological effect on the disease in question. If the results from the preclinical studies are favorable, then the compound is sent to human clinical trials, which are notoriously expensive to complete (hundreds of millions to billions of dollars) and can take up to 15 years to result in a commercially successful drug.
This is where the AI comes in. It doesn’t have to complete the entire work of the scientist, but rather to assist the scientist in their work on a daily basis. This will allow the scientist to complete their work much faster than before and enable them to focus on more important issues. For example, the AI can screen huge amounts of data in a matter of minutes and even make predictions about how molecules interact with each other and design new ones. All of these tasks would take the scientist years to complete on their own. In the end, without the AI the scientist would be missing out on a lot of potential and would also be working at a slow pace to find cures for the world’s diseases.
What Exactly is an AI Drug Discovery Platform?
AI drug discovery platforms are software systems for the various steps in the process of drug discovery. They are supported by artificial intelligence (AI), machine learning (ML) and deep learning (DL). Such a platform is not a single tool but consists of a variety of different tools which are combined in a system to support the daily work of the scientist in an efficient manner. In order to be efficient, the system has to be integrated in the workflow of the user. A large data set is needed as input for the system in order to train the models which are used for making predictions. Also the system has to be able to process a vast amount of data in a short period of time in order to be able to provide fast results to the user.
Imagine having a super-smart assistant that can read every scientific article ever written, every chemical ever synthesized and predict how it will behave in any biological system. This is what an AI drug discovery platform can do for you and your research. The AI platform processes your raw data (genomic sequences, 3D structures of proteins, large collections of chemicals, patient data, etc.) and turns it into powerful insights for the researcher and drug developer. These insights can aid in identifying new targets for drugs of action, designing novel chemical entities, predicting potential toxicity of candidate compounds and even identifying existing drugs to repurpose them for treating diseases of interest to you. This is not to say that the human scientist is any less important. However, the amount of data being generated in the life sciences today is exploding. That means there is a huge need for an intelligent tool to parse through all of this information quickly to draw meaningful conclusions to aid in the arduous drug discovery process.
From Target Identification to Clinical Trials: AI’s Role
There are many stages involved in the process of drug discovery, and in every one of them, AI can make improvements. But the greatest value of AI in the process of drug discovery is realized when it is used in all of the following stages:
Target Identification: Early Drug Discovery with AI. When researchers first begin to investigate a disease for which they wish to develop new treatments, they typically start with a hypothesis regarding the cause of the disease. They test this hypothesis by screening for modulators of protein activity, for example. However, the initial phase of target identification can be long and arduous. To identify targets for diseases using AI, large amounts of genomic and proteomic data are analyzed to identify novel targets and their corresponding biomarkers. AI systems can also identify potential targets within a large biological network and can even predict whether a potential target is likely to be druggable.
The next phase of the discovery process following the identification of a target is the Lead Discovery and Optimization process where the lead molecule(s) that interact with the target in a biological system are identified and then the properties of said molecule(s) are optimized. Most compounds that are screened by a high-throughput screening (HTS) assay do not interact with the target at useful concentrations. Those that do act as the initial ‘leads’ and the properties of these leads can then be optimized using AI to increase their potency and selectivity for the target in question. The AI can also design novel compounds that have the desired properties to act as effective drugs and the AI can design a cost effective synthesis route for said novel compounds. This makes the use of AI extremely powerful in the initial stages of the drug discovery process allowing researchers to make extremely rapid progress in the search for effective drugs.
The Data Deluge: AI’s Secret Weapon
A huge amount of data in the fields of biology and chemistry is being generated on an unprecedented scale. A huge amount of this information is published in papers that are freely available online, yet it is so vast that it is becoming increasingly difficult for individual researchers to sift through the information in order to be able to complete their work. This is where the data becomes a researcher’s secret weapon in the world of drug discovery.
This is where the AI truly has the edge. Because of the amount of data being generated by biology and chemistry today, the amount of data that a single human can process is dwarfed by what a computer can process, especially when using AI to process the data. In addition, when a computer is programmed with a deep learning model it can recognize patterns within data that would be impossible for any human to recognize, and the model can even learn from past experiments to make new predictions and even come up with new hypotheses. This is where the true potential of any good AI drug discovery platform is unleashed.
Choosing the right Platform for a Biotech/Pharmaceutical Company in the USA.
Also of interest for the manager of the Biotech and/or Pharma company from the USA are the platforms which already have been presented in this article and which I tried to describe in as much detail as possible as they already are developed and already have been put into production. All of them can be of interest for their own scope of application. If I would have to pick one of them for a large Pharma company I would not choose the first one since it seems more to be designed for the Biotech companies in the USA and every single Biotech company as well as every single Pharma company in the USA already has their very own ‘ecosystem’ and needs tools that can integrate perfectly with the processes that already are in place within a company, which also needs to be able to scale with the growth of a company and most important of all; has to deliver excellent results.
The best way for a company to select a suitable platform is for them to look beyond the marketing hype, features and functions of various solutions and view how the platform could fit into the current workflow of the researcher. It is also very important that the platform will grow with the company and deliver high quality results that can be used by the scientists to progress their research. Clearly the platform will be a significant investment for the company and therefore it is very important to get the correct return on investment. As such as getting the platform to meet the current needs of the company and in addition enable the company to meet their long term strategic goals.
Key Features to Prioritize
When choosing an AI Drug Discovery platform for your company there are a number of key features that you should consider non-negotiable. We have listed a few below.
Data Integration: How does the Platform Integrate with all of your Proprietary Data & also pull in all of the Public Data as well. As a biotechnology or pharmaceutical company, you most likely have a lot of proprietary data generated in-house, as well as experimental data from publications and public databases. The AI drug discovery platform you are considering to purchase should seamlessly integrate with all of your in-house databases as well as be able to import and pull in the external, public data as well.
Next is the Predictive Accuracy and Robustness of the models. The platform must be able to correctly predict molecular properties of small molecules such as solubility, permeability, lipophilicity, as well as the affinity to the target protein. The algorithms must also be able to predict toxicity. It is very important to validate the results of the models on existing data sets. Ideally, the platform should also provide case studies of how the platform was used to support the discovery of a molecule. It is also very helpful to run own data through the platform to get a feeling how the models perform.
3. Flexibility: To make the best use of any drug discovery program, hypothesis generation is key. Thus, it is extremely important that a platform is easily extendible by new models as well as by modifications of already existing ones. This especially holds for already complex diseases where the scientific community is still in the process of research and therefore models and compounds under investigation are subject to change often.
4) User-Friendliness and Support: This is often difficult to determine prior to purchasing. Be sure that the interface to the platform and its various tools are easy for your staff to use. Also, review the technical support provided by the vendor for the platform and its various tools, as well as any training or ongoing maintenance that will be provided to your end users. In AI Drug Discovery there are many complex tools and techniques being used and it is very important to have good support and training to ensure that they are being used correctly.
The “USA” Factor: What Biotech and Pharma Companies Here Need
The “USA” Factor: What Biotech and Pharma Companies Here Need
First, compliance with laws such as the Health Data Storage Act and the FDA guidelines for computerized systems such as in vitro diagnostic aids, which are for example required for regulated clinical trials in the USA. Next to being compliant from the start the chosen drug discovery platform has to be able to support all future IT requirements and in addition to support future data generation also be able to process the existing data in order to be able to submit it to the FDA for approval. We have seen many biotech companies being “stuck” for months because their old system did not support the required data integrity, traceability or documentation capabilities.
Next up is the all important point of Intellectual Property (IP) Protection offered by a platform for AI Drug Discovery. Given that the vast majority of data as well as the results from such research is of a highly proprietary nature it is of the utmost importance to ascertain whether or not the platform at hand will safeguard all data in the highest possible manner and protect said data from any potential misappropriation. In this context it is equally important to review the data ownership policy in order to confirm that your company will retain 100% ownership of all data as well as of all results from research generated while using the platform.
In addition to all of the above factors, it is very important to look for a platform that is well connected to other potential users. Such platform could be research hub, biotech or pharmaceutical company based near other similar organizations or university with strong research focus in the relevant area. Such platforms typically provide means for collaboration between users. Also, it is very important to have access to local talent pool that can provide necessary training and consultation services. As the center of innovation in the world, with numerous research hubs and biotech clusters, US provides significant advantages to any organization looking to utilize AI for drug discovery when compared to rest of the world.
Real-World Impact: Success Stories
Of the smaller to medium-sized biotechs that have put the platforms to use, lead optimization time has gone from months to weeks. A rare disease target had proven to be very difficult for a company to develop potent inhibitors for. After getting the AI-based drug discovery platform integrated into their workflow, they identified in a matter of weeks novel chemical scaffolds that had not been identified previously by the company using more conventional methods. There is a lot of potential for this technology to drive further innovation in the field of drug discovery but it is not yet ready for prime time to start to actually develop potential medicines.
FAQ
Is an AI drug discovery platform going to replace my scientists?
No. The intent of using AI drug discovery platforms is to assist the scientists and to greatly facilitate and to greatly assist the scientists in their daily work. And to greatly assist them in identifying new potential drug candidates. And to greatly facilitate the work of the scientist in the lead optimization process. And to greatly assist the scientist in designing new assays and new compounds to test for potential drugs.
How long does it take to implement an AI Drug Discovery Platform?
A basic integration may take a few weeks to a few months of months but a full integration of a AI platform with all of your current data, your current workflows, etc. and training of your staff can take 3 to 6 months.
What kind of data do you need to get started with an AI platform?
Generally, the more data that is available to a platform, the better it will be at identifying the optimal compound. Data that would be particularly useful include: 1) chemical structures (small molecules, proteins, etc.), 2) biological activity (assay) data, 3) genomic information (DNA, RNA, etc.), 4) protein structure data, and 5) preclinical and clinical data for compounds (if available). Most platforms are able to also access public databases to gain additional information.
Are these platforms secure. What about my proprietary data?
What about security? How does AI2ME ensure that my data (especially proprietary) is safe and sound? A: Security of your data is top priority for us. We use the strongest available encryption (e.g. AES) and implement strict access controls. We can set up your platform on your servers (on premise) or in your private cloud. We encourage you to review our data security policies and procedures in detail to make sure that they comply with your company’s compliance policies and IP protection strategy.
What’s the typical ROI for investing in an AI drug discovery platforms.
The ROI for using AI in the drug discovery process can be difficult to give a specific number for since most of the information that companies use AI for to find potential drugs are usually proprietary. But by cutting down the time that is spent for R&D and lowering the costs for R&D a company can save hundreds of millions to billions of dollars in a few years and make lots of money with potential drugs.
Can these platforms help with drug repurposing?
Yes, absolutely! While there are no guarantees of success when using AI to search for new indications for existing drugs, the tool can identify leads for potential drugs within a matter of weeks and save companies millions of dollars in the research and development process. The way the AI tool works is to analyze the molecular structure of a drug currently on the market, as well as the disease pathways affected by that drug. It can also look at the side effects of the currently marketed drug and use that information to suggest other potential uses for the compound.
What are the biggest challenges for implementing an AI drug discovery platform?
A) Data quality and standardization, integration with legacy systems, skills to effectively using the AI platform, change management and internal buy-in.
Conclusion
The landscape of drug discovery is rapidly changing and AI is no doubt going to play a very significant role. Thus, the use of an AI drug discovery platform has moved from being a ‘nice to have’ tool for the most innovative of US biotech and pharma companies, to being a necessary tool for them to stay in the game in terms of R&D. The software can enable faster drug discovery and development by for example identifying the very best drug targets, and then use said targets to optimize lead compounds to name but a couple of areas where the platform can add significant value. Subsequently the AI system can also assist in the complex process of finding the right dose of any drug or combination of drugs that are to be used for the successful treatment and/or management of serious human diseases.
In selecting the best AI Drug Discovery Platform for the US-based Biotech and Pharmaceutical companies, one needs to consider more than just the features and functionality of a platform. The platform must operate within the given company’s regulatory environment and above all, protect the Company’s valuable Intellectual Property. The platform should enable the Biotech/Pharma Company to leverage resources of the vast US scientific community to develop new, innovative and effective drugs for various diseases. We at Software Development Times can help you in selecting the most suitable AI platform for your company’s needs.