The slow and expensive process of bringing a novel, small molecule drug to patients begins at the earliest stages of discovery, where computational and medicinal chemists will spend months screening compounds, optimizing candidate molecules, and validating them to choose promising ones to advance through preclinical testing based on multidimensional parameters.
In November of 2021, Simulations Plus unveiled its Artificial Intelligence-driven Drug Discovery (AIDD) module to dramatically cut costs and reduce the time to identify candidate drugs. AIDD leverages two of the company’s core software products—its top-ranked, fully validated ADMET Predictor®, which takes molecular structures as inputs and uses machine learning technology to predict different properties for those molecules, in combination with its widely-used and flagship GastroPlus® software, which simulates the behavior of compounds in virtual animal and human models to predict absorption and systemic exposure.
“We have worked on these two separate approaches for many years. They’ve been independently validated to be among the most accurate available for predicting different physicochemical properties and simulating exposure around the body,” said John DiBella, president of the Lancaster Division of Simulations Plus. “We’ve now combined them in AIDD, so that researchers and scientists, at an early stage, can begin to rapidly optimize their molecules to not only be active for their target of interest but also to meet preclinical or clinical PK endpoints. That’s something no other tool on the market can do today.”
Remaking drug discovery
A chemist with limited software experience and a standard laptop computer can run the AIDD module using the intuitive, streamlined interface. No coding or special training is required. It’s designed to allow someone to begin generating molecules on the first day of use. All a chemist needs to do is input a seed molecule as a starting point and AIDD will continue to generate and test iterations at a rapid pace.
Instead of having chemists manually make modifications to candidate molecules to see which best achieves a desired set of endpoints, AIDD automatically does that with astounding speed and scores each molecule’s properties and activity. Using the software, a computational chemist can generate and test as many as 10 million molecules in 24 hours and rank them across a set of parameters relating to how they would behave as drugs in the body, such as their dissolution, absorption, and bioavailability. This allows chemists and DMPK scientists to optimize molecules across multiple dimensions and creates a ‘feedback loop’ or ‘2-way bridge’ between discovery and early development, accelerating drug discovery and increasing the chance of success in later stages of development.
As an off-the-shelf product, AIDD offers a compelling value proposition at the cost of synthesizing and screening a handful of molecules. “There are people who’ve done enough medicinal chemistry that they have good intuition and they’ll be able to guide that design process pretty well,” said Senior Director of Business Development Eric Jamois, for Simulations Plus. “These scientists can do that through sheer intuition and experience, but it takes quite a bit of time. Our software implements standard and custom rules that can be deployed consistently and at speeds that are orders of magnitude beyond what can be done in a conventional approach.”
The AIDD software is so user-friendly and intuitive that it is accessible not only to experienced medicinal chemists, but also to any translational researcher or drug discovery scientist. “I wish I had had this platform when I was developing my prostate cancer drug,” says Dr. Jones. “It would have halved my development timeline and cost.”
Finding a needle in a haystack in a millisecond
In August of 2021, Simulations Plus reported on the second phase of a collaborative research agreement with a large pharmaceutical company to evaluate the impact of the AIDD module. One of the main objectives of the collaboration was to design new molecules that had an activity below 50 nanomolar or for a specific target in the company’s therapeutic program of interest. Of the 19 molecules synthesized in the second phase of the collaboration, nearly 30 percent of the molecules met that goal, with potencies as low as 16 nanomolar.
“The average predictive errors for rat and human microsomal clearance, estimated using the ADMET Predictor default metabolism models, were impressive, illustrating the potential for integrating physiologically-based pharmacokinetic (PBPK) modeling within the AIDD module and reliably applying systemic exposure considerations during lead compound selection,” said Senior Principal Scientist Michael Lawless, for Simulations Plus. The company and its pharma partner expect to publish details of this validation of the AIDD module in the coming months.
One challenge chemists face in optimizing a molecule is that they typically do this for an individual property at a time. That can improve the performance in one aspect of a molecule but may sacrifice other properties in the process. AIDD performs its optimization all at once within the environment of a virtual animal or human model. This enables it to consider all competing factors that will be necessary to get a compound into the bloodstream and to its desired target at therapeutic levels while minimizing toxicity.
Another consideration for many drug developers during the discovery process is ensuring that selected compounds can be synthesized with ease, so they don’t require a significant investment of time, energy, or resources to produce. AIDD also automates this process and can evaluate the synthetic viability of a compound. This creates savings in the CRO budget by allowing drug developers to select a narrower set of compounds to take forward.
Plays well with others
That’s not to say that AIDD provides an alternative to internal AI groups. Instead, it’s an opportunity for companies to invest their time into designing modules that work with AIDD to address things it doesn’t do or do things it would prefer to do differently. The module is designed to be easily integrated with other off-the-shelf applications or custom designs built within a company.
Users can build APIs to link their own databases or develop workflows to meet their own specification. They can even use their own familiar graphical user interface if they like. “Instead of using ours, you can call an external program now within AIDD,” said Simulations Plus’ Jamois. “Companies that have scientists coding AI solutions can use them to do bits and pieces we don’t do.”
Simulations Plus prides itself on providing a high level of customer service, building additional global or local models based on a client’s unique datasets, and working closely with them to ensure such integrations run smoothly. In fact, Simulations Plus also offers it’s AIDD capabilities as a service. If a company doesn’t have the time or resources for running the AIDD software itself, Simulations Plus can do the work for them. “Hand us a list of hits from your screen or a lead series, and we’ll give you back a curated list of compounds with high predicted activity and ideal PK/PD properties. We’ll even make sure the compounds are in novel chemical space and order them for you,” says Jones.
“We’re not seeking to eliminate the human element from the drug discovery process or the need to conduct web lab experiments, but rather enhance the process and make it more efficient,” said Jamois. “Our results with AIDD testifies to the powerful combination of human interaction with artificial intelligence to generate new ideas, test hypotheses, and optimize molecules to meet safety and preclinical/clinical efficacy endpoints in the clinic. As a result, it bridges discovery and early development and avoids costly failures later in the process.”
To learn more about AIDD module and what it can do to accelerate the drug discovery process for you, contact us here.