Drug Discovery & Development

Insight from TileDB's Stavros Papadopoulos on why pharma needs to pump the brakes on AI hype and get their data house in order first: "Don't touch AI without data infrastructure," he said in a recent interview. While Deloitte projects pharma could save $5-7B over five years with AI, here's the reality check: Only 16% of drug discovery efforts currently use AI (Deloitte estimate from earlier this year) Data scientists spend roughly 80% of time just preparing data (estimates vary but this is a commonly cited ballpark range.) Much life science data — maybe on the order of 99% — isn't tabular, yet many many data science tools are built for tabular data. Papadopoulos' self-described "radical" but practical view: 1. Build robust data infrastructure first before "touching AI" 2. Understand fundamentally that there is no such thing as "unstructured data" -- even white noise and ciphertext has a structure. 3. Papadopoulos goes so far as to say that there are no best practices when it comes to data management in life sciences — given the relative immaturity of the field. But he then goes on to recommend what best practices would look like conceptually and notes what it would take to get there. The answer isn't rushing to shoehorn AI into everything at the outset. While tools like LLMs can accelerate some rote tasks in short run, that's not going to lead to any kind of meaningful 'digital transformation.' Sometimes a calculator beats an LLM — in performance and certainly cost. #LifeSciences #AI #DataScience #Pharma #Innovation #RD

Eli Lilly and Company just announced data bolstering the case that tirzepatide is more effective than semaglutide in supporting weight in a ph3b trial. Had a look to see if any of the prior announcements appeared to have a correlation with stock performance. Not really.... Here's a recap of some of their clinical announcements over the past year juxtaposed with stock performance Individual Announcement Impact Strongest positive reaction: SUMMIT Trial Results (+3.50% day-of return) Most significant downside: SURMOUNT-1 Results (-6.53% 5-day return) Highest volatility: SUMMIT Trial (2.11x normal volatility) Most stable: Donanemab Approval (0.51x normal volatility) Statistical Significance None of the individual announcements showed statistically significant price impact (all p-values > 0.05)

Things are looking up for Eli Lilly and Company A phase 3b open-label randomized clinical trial (SURMOUNT-5) has shown that tirzepatide leads to greater weight reduction compared to semaglutide in adults with obesity or overweight conditions. The study from Eli Lilly demonstrated a mean weight loss of 20.2% with tirzepatide (Zepbound) versus 13.7% with semaglutide (Wegovy) over a 72-week period.

Which major #pharmaceutical company leads the way in research and development? View this year's Pharma 50 rankings: https://lnkd.in/evneuC2Y

Which companies had the most academic citations from 2019 to 2024 (as of November)? Merck and Microsoft came in at the top. The top 10 were all tech and pharma firms. Pfizer, China-based Tencent and Bristol Myers Squibb rounded out the top 5.

Can off-the-shelf genAI be trusted for scientific research in fields like drug development? Oftentimes no. Which is why I enjoyed chatting with Joseph Mullen, director of data science and professional services for SciBite at Elsevier about the need for sophisticated retrieval augmented generation (RAG)-based systems (including agentic RAG!) "You need explainability and the ability to validate hypotheses to avoid critical errors," Mullen said. There are two aspects here. One, "It's important that humans remain at the center of decision-making," Mullen said. But the other aspect is the agentic RAG one -- designing AI systems to take more accountability for their accuracy and ability to accomplish a defined scientific objective. “The information retrieval step could involve a combination of ontology-based and vector-based methods,” Mullen explained. “This means recognizing that questions can go to multiple sources, determining optimal routing pathways, ranking returned information, and executing complex process chains.” Full article in the comments.

Just published: The fully upgraded weekly AI research roundup Explore super-comprehensive weekly AI research roundup thanks in large part to a collaboration with Frédéric Célerse, PhD. For the full overhauled roundup, check out the link in the comments. Drug Discovery News Insilico Medicine Secures 10th FDA IND Clearance Insilico Medicine has achieved its 10th FDA Investigational New Drug (IND) clearance for ISM5939, an AI-driven cancer drug developed in just three months using their proprietary PharmaAI platform shares https://lnkd.in/ezm33_2n Enveda Raises $130M Series C Recursion and Exscientia Complete Merger The strategic merger between Recursion Pharmaceuticals and Exscientia has been finalized, creating a powerhouse with 10 drug candidates in the pipeline and a capital runway secured through 2027. https://lnkd.in/edReZ7nc AstraZeneca Partners with Lunit Oncology Research Infrastructure Mount Sinai Health System Launches 65,000 sq ft AI Center Amazon Unveils "Olympus" Multimodal AI Model Technical Advances Hugging Face has launched SmolVLM, a 2B-parameter vision language model optimized for on-device inference. HF's Philipp Schmid also has the scoop on how we may be able to push LLM optimizations farther than expected. "A new study achieves 98% accuracy recovery on key benchmarks while removing 50% of Llama 3.1 8B's parameters using pruning," he wrote. DINO-X Sets New Object Detection Benchmarks OneDiffusion: Unified Generative and Predictive Model Marco-o1 Enhances Open-Ended Reasoning in AI Systems Global Moves Mistral AI Expands to Silicon Valley French AI startup Mistral AI, valued at €6 billion, is establishing operations in Palo Alto, California. OpenAI Establishes Paris Office South Korea Launches AI Safety Institute Additional Highlights Anthropic Releases Model Context Protocol (MCP) Commonwealth Bank of Australia Enhances Customer Service with AI AI-Generated Content Dominates Long-Form Posts According to Originality ai, 54% of English-language posts on LinkedIn exceeding 100 words are now AI-generated. In a recent TED talk, Max Jaderberg, CEO of Isomorphic Labs, discussed how AI is transforming scientific research through "AI analogs" of biological systems. "AI in drug discovery might not yet live up to the industry hype, but that doesn’t mean the future isn’t bright," wrote DeepMirror cofounder Max Jakobs in a summary of a post on what's going on with AI in drug discovery: https://lnkd.in/en-dZ_4V #AIResearch #Biotech #ComputationalChemistry #TechNews #DrugDiscovery #AIInfrastructure #GlobalAI #TechnicalAdvances #AIIntegration #MachineLearning #AIInnovation

Novartis, Johnson & Johnson Innovative Medicine, Bristol Myers Squibb, Roche and Amgen appear to be the overall patent winners of 2024 (based on international patents granted) based on data pulled today. Yes, the year isn't over yet, but the companies have a strong lead thus far. For more data, check out the LinkedIn post here. Thanks to Maryam Daneshpour for inspiring the most recent research providing some input on what to look for when querying patent data. More data coming!

Two GLP-1 pioneers have surpassed other Big Pharmas in terms of market cap. Data from November 26 - via yfinance.

A Geographic Analysis of Pharma Innovation Centers Maryam Daneshpour and I are teaming up to cover the concentration of pharma patent activity (with financials) to see where breakthrough research is happening. The first map shows total patent concentrations weighted by total patents in the first ten months of 2024 weighted evenly (50/50) with market cap. - Unsurprisingly, the Boston-Cambridge corridor shows intense activity, with companies like Moderna and Biogen showing high R&D efficiency despite their smaller size - The San Francisco Bay Area cluster reveals robust patent output relative to R&D investment - Switzerland's Basel region shows how geographic proximity between Novartis and Roche continues to drive R&D and financial output. - The German research hub including Frankfurt, Heidelberg, Mannheim (and Darmstadt shows resilience 1. Boston: Research, Biotech, Academia 2. NJ-NYC: Big Pharma, Manufacturing 3. Basel: Big Pharma, Research 4. SF-Bay: Biotech, Startups 5. Shanghai: Manufacturing, Research 6. Tokyo: Research, Manufacturing 7. San Diego: Biotech, Research 8. Philadelphia: Research, Manufacturing 9. Beijing: Research, Development 10. Paris: Research, Manufacturing 11. Darmstadt: Manufacturing, Research 12. London: Research, Biotech 13. Seattle: Biotech, Research 14. Research Triangle: Research, Manufacturing 15. Copenhagen: Research, Biotech 16. Munich: Research, Development 17. Singapore: Manufacturing, Research 18. Dublin: Manufacturing, Development 19. Tel Aviv: Research, Startups 20. Melbourne: Research, Development