AI-powered platform accelerates drug discovery from molecules to medicines

Every step forward in drug discovery depends on scientific decisions that can determine whether a promising therapy advances, stalls in development or never reaches patients who need it most. As pharmaceutical companies increasingly use artificial intelligence and machine learning to support research, the challenge is no longer just building models. It is turning these models into reliable tools that scientists can use throughout the drug discovery process.

For Dompé farmaceutici, addressing that challenge means expanding how it uses AI in R&D. - Founded as Milan’s first compounding pharmacy, Dompé has grown over 130 years into a global biopharmaceutical company, specializing in both primary and specialty care and biotech, with a particular focus on ophthalmic, neurological and pain-related conditions.

Through its Exscalate fully integrated drug discovery platform, Dompé combines advanced analytics and AI to help researchers analyze vast volumes of biological, chemical and clinical data accelerating design-make-test-learn (DMTL) cycles. The platform identifies potential therapies, predicts how they may perform and helps determine which patients are most likely to benefit.

According to Andrea Beccari, Head of Discovery Platform and Vice President at Dompé, medicine has increasingly shifted toward a predictive approach in recent years. While significant progress has been made, many of today’s greatest health care challenges involve complex diseases such as cancer, cardiovascular conditions and immune disorders. Because these diseases are driven by interconnected biological factors, they often cannot be effectively addressed by targeting a single cause.

Advancing treatment development, therefore, requires a deeper understanding of disease complexity. This begins with the collection of the right data to build more accurate models and generate better insights.

To support this effort, Dompé turned to SAS Viya.

Choosing a platform to operationalize AI at scale

According to Beccari, the company chose SAS Viya to address two priorities:

• Build an enterprise-grade analytics platform for continuous integration and deployment of Exscalate’s computational drug discovery models.
• Identify a technology partner that could help operationalize advanced analytics at scale.

“The quality of the SAS team – their ability to co-design and implement solutions alongside our Exscalate data analytics pipeline – was a decisive factor,” he says.

Beccari also points to the platform’s capabilities as a key differentiator: “SAS’ combination of trusted, validated AI and machine learning models and seamless enterprise integration would be hard to replicate. No other platform we’ve evaluated reaches the same level of trust, modeling quality and integration.”

Today, SAS serves as a foundational component of the Exscalate ecosystem.

“SAS is a strategic partner delivering both scientific and operational impact,” adds Beccari. “The SAS team is deeply integrated with the Exscalate team, effectively insourcing specialized competencies and capabilities that strengthen our platform.”

Creating an AI factory for modern drug discovery

As Dompé’s predictive models became more sophisticated, the company needed a way to industrialize how those models were developed, validated and deployed across the organization.

Working with SAS, it created an AI factory within Exscalate that transforms large-scale experimental data into production-ready analytics. At the center of this capability is ProfhEX, Dompé’s platform for small-molecule polypharmacology profiling. The platform ingests millions of data points, automatically generates and validates models, and delivers thousands of APIs that are immediately available to both researchers and AI agents.

The ProfhEX polypharmacology models illustrate this scale: The latest release covers close to a thousand predictive models across roughly 690 human targets, trained on more than 5 million curated bioactivity measurements, with profiling extended to ADMET and safety endpoints.

Because every model is produced through the same governed pipeline, each prediction is delivered with its reliability context: uncertainty estimates and an applicability domain that tells researchers when a compound falls outside the chemical space the model has actually learned. That context allows the output to be used as decision support rather than as an unqualified number.

Accelerating model deployment

One of the most significant benefits Dompé has realized from adopting SAS Viya is a dramatic reduction in the time required to move analytical models from concept to production.

“SAS Viya significantly reduced the time it takes to move from an initial prototype through model development and validation to full production deployment,” says Beccari. “What once took weeks or months now takes just hours or days.”

The ProfhEX platform has transformed the DMTL cycle, allowing researchers to focus more on scientific design and decision-making rather than manual data preparation, model development and validation.

In addition, Dompé uses predictive analytics and AI to evaluate potential toxicity and side effects earlier in the development process, helping researchers identify risks before therapies reach later stages of development.

“By doing predictive analysis earlier in the cycle, we can speed our drug development process and deliver solutions to patients faster,” says Beccari.

Building trust in AI-driven research

In pharmaceutical research, speed alone is not enough. Every model that informs scientific or business decisions must be explainable, auditable and validated.

“It means setting rigorous model quality acceptance criteria upfront, applying extensive validation procedures and maintaining a fully auditable process from data ingestion all the way through model deployment,” says Beccari.

SAS Viya helps make the analytics life cycle reproducible and traceable end to end. Combined with prevalidated model templates and strong governance, it gives Dompé confidence in increasingly sophisticated AI and machine learning models.

Validation is especially critical in pharmaceutical research. “When defining models for predictive analytics, there is always a risk of inconsistency between model development and application,” says Beccari. “The model must make sense not only statistically, but also in terms of how it will be used in practice.”

Balancing innovation with governance is central to Dompé’s development process. The company rapidly prototypes and selects AI and machine learning approaches for specific research goals, then works with SAS to turn those prototypes into a production-ready solutions within a governed framework. This collaboration also enables Dompé to incorporate advanced methods, including Bayesian neural networks, to better assess model uncertainty.

Supporting collaboration between people and AI

As AI becomes more deeply embedded in research workflows, Dompé sees analytics platforms evolving into shared environments where humans and intelligent systems collaborate.

Using SAS Viya, models can be deployed as standardized APIs and accessed consistently across research activities. Whether a scientist is exploring a new hypothesis or an AI agent is executing a workflow, both rely on the same trusted analytics.

SAS also helps connect the teams behind Exscalate’s research, platform development and model deployment. Drug discovery teams rely on up-to-date predictions to guide decisions, while technical teams build and operationalize those models through the Exscalate AI factory. “SAS is the engine of that factory,” says Beccari.

Advancing the future of drug discovery

Dompé sees advanced analytics and AI playing an even greater role in the future of drug discovery.

“AI-empowered data analytics and computational simulations will be central to accelerating and strengthening our decision-making process across the entire R&D pipeline,” says Beccari. “Looking ahead, the company envisions deeper integration of AI agents that use model APIs and contextual knowledge to support researchers in real time.”