[Adobe Stock]
Looking ahead to 2025, two execs predict more of an evolutionary advance of digital initiatives in drug development than a revolutionary one.
1. AI pivots to biology-first approaches
BPGbio
BPGbio CEO Niven Narain predicts a pronounced shift in how the industry applies AI: “The biopharmaceutical industry will pivot from hype-driven investments in AI to a focus on biology-first AI approaches that prioritize real longitudinal biological data.” This transformation will emphasize integrating AI with patient-derived multi-omics datasets, enabling deeper understanding of disease mechanisms and therapeutic responses. The goal? Moving beyond correlation to uncover causation in drug discovery, particularly for complex diseases like glioblastoma and pancreatic cancer. Narain points to the promise of marrying AI with rigorous biological inputs and translational models, which will help the sector address unmet medical needs.
2. Steady growth in digital twin adoption
Gen Li, Ph.D., CEO of Phesi, sees a gradual but transformative adoption curve ahead: “People are going to become more interested in digital twin adoption as education and training efforts continue and as understanding and awareness of the area improves. It will initially be a gradual process, but at some point it’s going to explode as people become fully familiar with the benefits. The industry needs to be ready to embrace that.”
3. A structured digital twin adoption pathway emerges
Phesi’s Li foresees a structured adoption pathway starting with Digital Patient Profiles (DPPs). “DPPs support key decision-making processes, allowing medical teams to assess whether there’s enough data for a digital twin to be part of their core strategy,” Li explained. These profiles enable early FDA engagement and improve trial design even when full digital twin implementation isn’t feasible.
Phesi
Digital twins are virtual representations of patients, constructed using significant amounts of historical and current clinical data. They simulate patient responses to treatments, potentially replacing or augmenting control arms in clinical trials. Yet building meaningful digital twins requires substantial volumes of high-quality data.
“Building digital twins for rare diseases can actually be challenging, since with many such indications, there is a lack of data in required volumes to make a digital twin meaningful,” Li noted. “However, in other areas where placebo groups can be ethically challenging, digital twins can be very effective—particularly in cancer, hematology, and neurodegenerative conditions.”
Li cites specific examples where digital twins have already shown promise:
- G12C KRAS Mutation Case Study: Despite no prior clinical trials investigating the G12C KRAS mutation, Phesi created a digital twin through the use of datasets from biomarker analyses. This approach allowed researchers to simulate patient responses and design more effective trials without starting from scratch.
- cGvHD Digital Twin for Prednisone Control Arms: In chronic graft versus host disease (cGvHD), Phesi developed a digital twin to replicate patients receiving prednisone, the standard-of-care treatment used in control arms. “This allows researchers to accelerate trials into this challenging condition by automating an external control arm and enabling sponsors to test new therapies faster,” Li explained.
- Alzheimer’s Trial Duration Reduction: In neurodegenerative diseases like Alzheimer’s, digital twins have demonstrated the potential to significantly reduce trial durations. “The standard duration for Alzheimer’s clinical trials is 18-24 months,” Li said. “By using digital twins, we’ve been able to model a shorter, 12-month observation period, during which we could observe a 20-30% rate of cognitive decline. This approach could lead to much shorter clinical trials with improved outcomes.”
4. Expect more AI partnerships and regulatory wins for biopharma in 2025
While there are yet to be any FDA approved drugs discovered solely with AI approaches, it’s only a matter of time before that milestone becomes a reality. In February 2023, Insilico Medicine’s lead candidate, INS018_055, a small molecule inhibitor for idiopathic pulmonary fibrosis, won the FDA’s first Orphan Drug Designation for an AI-discovered and designed drug. The number of partnerships between pharma heavyweights and AI focused drug discovery and development firms is also increasing. For instance, Schrödinger, known for its physics-based computational platform, entered a multi-year collaboration with Novartis in November 2024. A number of company are already announcing wins. For instance, GSK accelerated its respiratory syncytial virus (RSV) drug trials by two years through real-time tracking of infectious diseases using predictive data modeling, according to its CEO Emma Walmsley, who was quoted in Barron’s. Expect more of this in 2025.
Filed Under: machine learning and AI
