By Deborah Borfitz 

June 3, 2026 | VeriSIM Life is building the “full-stack predictive infrastructure” to help pharma companies foresee the challenges ahead in translating a molecule from laboratory discovery to a successful human therapy. The help comes in the form of a dynamic, “credit-score-like” assessment of the risk with inherent explainability and trust, according to Jo Varshney, DVM, Ph.D., founder and CEO of VeriSIM Life. 

Editor's Note: This article was originally published on Bio-IT World News. To read the original article, visit here.

As is well known in the industry, only about 1 in 10 drug candidates that enter clinical trials make it to market approval. The VeriSIM Life tool, part of the company’s BIOiSIM drug discovery platform, is designed to assist drug developers in making better decisions early in the high-risk business of medical research, Varshney says. It literally flips the stat by achieving nearly 90% accuracy in predicting clinical trial success across validated retrospective and prospective studies, helping drug developers make earlier, evidence-based decisions. 

“Predictions are made using hybrid AI models, which combine virtual animal-to-human drug simulations with machine learning,” says Varshney. “The system models how a drug behaves and explains the underlying biological and physical reasons for that behavior. The approach blends the cause-and-effect understanding of algorithms that follow scientific principles with those that can recognize patterns when analyzing vast datasets.” 

The key challenge facing pharma companies individually and collectively “ultimately comes down to translatability,” says Varshney. While the volume of predictions has reached an unprecedented scale, the field has shifted its focus from quantity to quality. Sponsors want predictions they can turn into decisions that hold up in the clinic. 

“VeriSIM Life, launched in 2017, spent its first five years rigorously validating BIOiSIM across disease areas, targets, partner programs, and internal drug development efforts,” Varshney says. It now has about 20 active partnerships with pharma and co-development partnerships, as well as a pipeline of its own assets “because we want to walk our talk.” 

One of those assets, a small molecule for pulmonary arterial hypertension, is less than a year shy of entering human trials. The company has already enabled several client programs to enter clinical trials significantly ahead of traditional schedules, she notes. “Probably we are the only company who has demonstrated this level of translational validation across both partner programs and internally developed assets.” 

Key Questions 

Most of the predictive analytic tools available to pharma companies focus on drug discovery, says Varshney. “The field has become increasingly good at generating molecules, but molecule generation is no longer the primary bottleneck. The hard question is which molecules are going to work safely and effectively in humans.” The holy grail is knowing which of those molecules can pass through clinical trials and become market approved.  

The questions VeriSIM Life typically hear from study sponsors are about which candidate molecules will work in patients or be best suited for a specific indication, what adverse effects can be anticipated and lessened during toxicology studies, and “ways to identify the right population to enroll in clinical trials to get the statistical significance and endpoints defining a successful outcome,” she says. 

Pointed single solutions exist that are specific to the chemistry of molecules, clinical trials, or patient populations, but they can’t predict the outcomes because they’re only focused on one of multiple problems that could emerge during human studies, continues Varshney. Biology is complicated, and the systems are fundamentally interconnected. 

Her company uses a combination of standard and proprietary machine learning and AI approaches, says Varshney, noting that the BIOiSIM multi model method blends diverse AI algorithms into a virtual drug development engine. “We look at the collection of AI models and mechanistic models we have to identify the best combinations that will help us solve the problem that we’re tackling.”  

Mechanistic simulations generate biologically grounded features that are integrated with hybrid AI and computational models to improve translational prediction robustness and explainability. 

VeriSIM Life aims to “connect different aspects of the translatability of a molecule”—including the chemistry, biology, and animal and patient variability—to provide meaningful, real-world predictions about the challenges a molecule is going to face in its drug development journey, Varshney says. From this understanding, a drug is assigned a “credit score,” which much like a person’s financial health dynamically changes based on the underlying data—e.g., modifications to its chemical structure, dosage, or the targeted disease. 

Limitations of the approach are tied to the availability of information, be it in the form of data or domain-specific human expertise, she continues. BIOiSIM identifies those gaps transparently so customers can understand both the confidence level and the boundaries of each prediction. 

Synthetic Data 

To enhance translation predictions, VeriSIM Life creates its own synthetic data through the BIOiSIM platform, to expand the rather limited external and lab datasets, reports Varshney. The company creates mathematical representations of human and animal systems to simulate how drugs interact with the body, which helps predict efficacy and toxicity in humans even when input data is limited. 

The platform learns from broad chemical and biological patterns across large datasets and applies those insights to specific drug candidates and disease contexts where experimental data may be limited. “From that lens, we have tested our system around 72 disease areas and thousands of different targets which connect to those areas,” she says. 

Uses Cases 

Without revealing company names or details, Varshney can share that VeriSIM Life was able to help one of its partners with an interesting target and multiple small molecules determine the best indications to go after from specific cancers to rare diseases. “We deployed our platform and the Translational Index to rank order the best disease areas that it should be focused on.” The customer validated the prediction and now has that program at the end of phase 1 clinical trials. 

In another partnership focused on the chronic toxicity of a drug candidate, VeriSIM Life is helping several companies reduce or replace the need for traditional, six- to nine-month studies in monkeys using AI-driven digital twins of both animals and humans to simulate how the molecule will behave in the body, she says. The work is directly tied to the Food and Drug Administration’s active transitioning away from animal testing by prioritizing new approach methodologies, including AI, computational modeling, and organ-on-a-chip technology. “We’re excited to be at the forefront of this implementation by the agency.” 

Varshney, a veterinarian by training, is understandably passionate about solving the translation problem in drug development in ways that reduce the need to involve animals. Her father was in the pharmaceutical business in India, so she learned early on in life how hard it is to bring a drug to the market. “But we can’t just keep saying that ... we have to dig deeper because ultimately ... getting even 30% of drugs approved is a massive win”—financially for companies as well as health-wise for patients. 

“We are truly gung-ho about solving translatability from preclinical to clinical,” she says, although it has admittedly been a hard journey. “I have spent the entirety of my career doing that, from understanding the problem set, building coding systems, understanding what that could look like from an AI perspective, and then starting VeriSIM.”  

Role of Organoids 

Like the BIOiSIM platform, microphysiological systems and organoids need to be human-relevant if they are to replace studies in animals, says Varshney. “They’re built to get closer to what’s seen in the patient population, but from a single point of view” because they’re primarily organ-specific and used to study specific cellular processes within that context. 

But the shift has begun toward simultaneous drug testing in different 3D organoids (e.g., brain, heart, and liver) microengineered to accurately replicate human physiological, functional, and pathological responses, she notes. Cell culture models and lower animal systems generally don’t have any human relevance. 

“You don’t have to be a veterinarian to appreciate the differences between horses, monkeys, dogs, and humans,” says Varshney. That mindset should extend to the translatability of drug testing across different animal models, so better preclinical and clinical trials are being built.  

Organoids are a scalable way to bridge the translational gap and provide high-quality, human-relevant data to train and test AI models, but they’re not easy to build, adds Varshney, whose Ph.D. research involved mini-organs. “If I had to recreate pharma from scratch, I would focus on creating 3D systems that are interconnected” to generate data that is structurally, functionally, and genetically much closer to actual human tissues than traditional 2D cell cultures and would also work toward eliminating animal testing where translational relevance to human biology is limited or non-predictive.” 

It’s a transition of consequence to the entirety of pharma and thus too big for any one company to take on, she adds. The beauty of the AI infrastructure of BIOiSIM is that it could take diverse datasets from these physiologically relevant 3D models to inform their use in better navigating molecules to the clinic. That shift could help the industry move from trial-and-error development toward a more predictive, human-relevant, and knowledge-driven model of drug translation.