Lilly's AI & ML Innovations Drive Breakthroughs in Healthcare R&D (2019-2025)

1. Summaries of AI/ML Initiatives by Period
– Q1 2021–Q4 2022 (Pipeline build-out and digital-health pilots)
• Connected-care devices – Initiated Ph 2 automated insulin-delivery studies combining continuous glucose monitors, smart pens and feedback algorithms to approach a closed-loop (“artificial pancreas”) system (Q2 2020).
• Digital diagnostics – Developed and launched a lab-developed blood-based p-tau217 assay and expanded PET imaging (Tauvid®) for Alzheimer’s to accelerate patient identification; in parallel built infrastructure to collect and analyze multimodal patient data.
• Protein-degradation and peptide discovery — In‐licensed Lycia’s degrader platform (Q2 2021) and invested in peptide triagonist design (GIP/GLP-1/glucagon) using in‐silico modeling.
• Connected-pen launch – Filed U.S. and EU regulatory submissions for a reusable digital insulin pen integrating dose capture, mobile-app connectivity and ML-driven dosing reminders (Q2 2019).

– Q4 2022 (AI partnerships and gene-therapy hub)
• OpenAI collaboration – Announced a multi-year research agreement with OpenAI to apply generative AI and large-language models for “novel-antimicrobial” discovery, aiming to accelerate hit identification and lead optimization (Q4 2022).
• Seaport Innovation Center – Opened a Boston-based hub for genetic-medicines R&D to leverage computational genomics, AI-driven vector design and high-throughput functional screening.
• Gateway Lab (U.K.) – Established a UK R&D lab focused on early-stage gene editing and bioinformatics collaboration across Europe.
• Morphic Therapeutics acquisition (Q4 2024) – Acquired an AI-driven small-molecule integrin-inhibitor platform to discover oral α4β7 agents in inflammatory bowel disease.

– Q1 2025 (Continued digital expansion)
• Pill-based incretin discovery – Advanced orforglipron, an oral small-molecule GLP-1 agonist discovered via computational chemistry, into Phase 3.
• ROI-driven analytics – Leveraged real-world data and ML models (e.g., Aon study) to quantify cardiovascular outcomes (MACE) impacts of obesity therapies, supporting payer discussions.

2. Comparative Analysis and Trends
– Shift from device-focused ML pilots (insulin pens, closed-loop pumps) toward AI-powered drug discovery collaborations (OpenAI for antimicrobials, Morphic small-molecule platform).
– Early-stage emphasis on digital therapeutics and biomarker analytics (p-tau217 assay, CGM algorithms) has grown into broad AI/ML integration in molecular design and high-throughput screening.
– Regulatory filings increasingly reference AI-enabled components (connected devices, digital diagnostics) alongside traditional pipeline milestones.
– Investment in gene-therapy and genetic-medicines centers signals a convergence of AI-driven bioinformatics with advanced delivery modalities.

3. Salient Observations
– Cross-Therapeutic-Area AI Deployment – AI/ML applied across diabetes (connected insulin delivery, digital pens), neuroscience (biomarker assays, disease-progression modeling), immunology (protein degraders, peptide design) and antimicrobials (generative models).
– External vs. Internal Innovation Balance – Increasing use of external AI partnerships (OpenAI, Lycia) to complement in-house computational chemistry, reflecting a hybrid R&D model.
– Data-Driven Payer Engagement – ML analyses of real-world evidence (e.g., Aon ROI study) strengthening economic value propositions for high-cost therapies.
– Regulatory & Commercial Complexity – Integration of AI in devices and diagnostics carries novel regulatory pathways and payer coverage challenges that require early engagement.

4. Explanations of Complex Elements
– Generative AI in Drug Discovery – Large-language and diffusion models can propose novel molecular scaffolds, predict binding affinities and suggest synthetic routes, reducing cycle-times from years to months.
– Closed-Loop Systems – Machine-learning algorithms trained on CGM and insulin-dose datasets can forecast glycemia and automate basal or bolus delivery, improving A1C and reducing hypoglycemia risk.
– Biomarker Modeling – Computational disease‐progression models (e.g., C‐PATH AD models) integrate demographic, imaging and fluid-biomarker data to predict placebo trajectories and amplify signal in clinical trials.

5. Conclusions
– Lilly has systematically embedded AI/ML across its value chain – from device design and digital therapeutics to diagnostics, biomarker development and drug discovery.
– Strategic external collaborations (OpenAI, AbCellera) amplify proprietary in-house computing platforms, accelerating novel antimicrobial and antibody programs.
– Targeted areas for AI/ML include diabetes (closed-loop insulin delivery, oral peptide design), neuroscience (digital and blood diagnostics, disease modeling), immunology (computational degrader and triagonist design) and precision oncology (retrospective biomarker analytics, small-molecule discovery).
– While regulatory and reimbursement pathways evolve to accommodate AI-enabled technologies, Lilly’s early investments position it to rapidly validate and commercialize next-generation therapies across multiple high-value disease areas.