Oracle Life Sciences’ Post

Want to leverage Ai in your drug discovery analysis? Then you need to have a solid data foundation in place. Before any meaningful AI analysis can occur, data must be in a machine-readable format that meets high-quality standards. This is one of the important benefits of a scientific data management platform. By ensuring data quality and structure, scientific data management platforms provide a solid foundation for predictive analytics, machine learning models, and other advanced data analysis techniques. https://lnkd.in/dUVUg44n

Great practical use case of leveraging AI in Life Sciences. 👌 🤓Learn more in the announcement below. 👇👇👇👇👇

For decades, researchers struggled with a lack of #data available in #clinicaltrials. Today we have the oppositive issue; an abundance of it which makes it difficult to organize and discern relevant from superfluous information. Organizations are now looking at #AI to address and streamline the quantity of information across varying sources. Learn more: https://bit.ly/3WeTqJO

Extracting information from #healthcare & #lifescience records as well as generating reports is saving time. Learn more about what Oracle offers such as #HealthNamedEntityRecognition, #RelationshipExtraction, #AssertionDetection & #MedicalEntityLinking #LLM #GenAI https://lnkd.in/gkEANrBK

Good tool to use in Applied Stats and Data Science is Bayesian Statistics. The methodology is applied as Bayesian research methods empower decision makers to discover what most likely works by putting new research findings in context of an existing evidence base (Mathematica, 2024) Ask AI: Clouds at sunrise and rain If there are clouds at sunrise, then it will rain on that day, and the probability is P(rain | clouds at sunrise). For example, if it rains 25% of the time and it's cloudy at sunrise 15% of the time, then 50% of the time it rains, clouds are at sunrise, and the probability is P(clouds at sunrise | rain) = 50%. #ML #Datascience #Bayesian #Tech #Systems #Bigdata #AI #OpenAI https://lnkd.in/gpdv9xAj

🌟 New Launch Alert We're thrilled to announce the launch of Dimensions Knowledge Graph, a game-changer for the pharmaceutical and life science sectors! 🎉 Unlock the power of internal and global research data with the Dimensions Knowledge Graph, powered by metaphacts GmbH. Dive into approximately 350 million records and 50+ public datasets, seamlessly integrating external knowledge with your internal data repositories. Curious to learn more about how Dimensions Knowledge Graph can revolutionize your research and development process? 🤔 Find out more in our news article https://lnkd.in/eUk4rHEk Discover how Dimensions Knowledge Graph can fast-track target discovery, streamline processes, and accelerate drug discovery. Don't miss out on this opportunity to supercharge your insights! #DimensionsKnowledgeGraph #AI #Research #DataIntegration #DrugDiscovery #LifeSciences

As efforts to enhance the robustness and reliability of #RWE in decision-making progress, deriving accurate #causalinferences from #RWD becomes increasingly crucial. Advanced methodologies, such as the targeted maximum likelihood estimator (TMLE) offering doubly robust #machinelearning-based methods, are being proposed to address the limitations of current methods. But what are the benefits of these new advanced methodologies, and how are they perceived by decision-makers? This important topic was discussed during the workshop at ISPOR—The Professional Society for Health Economics and Outcomes Research 2024 in the session, ‘Targeted learning for causal inference using real-world data.’ Chair Suzanne M. (Medlior Health Outcomes Research Ltd.) was joined by her colleague John Paul Ekwaru, Mark van der Laan (University of California, Berkeley) and Stephen J. Duffield (NICE - National Institute for Health and Care Excellence). In this Deep Dive from The Evidence Base®, we recap the key takeaways and discussions from the presentations. #heor #healtheconomics #outcomesresearch #realworldevidence #realworlddata #marketaccess #pharma #biopharma #medicalaffairs #regulation #regulatoryaffairs #hta #healthtechnologyassessment #healthdata #healthequity #healthpolicy #AI #artificialintelligence #ISPORAnnual

CREATE and the PHRI Population Health Research Institute have been collaborating for several years on how we can use #AI to improve the quality and efficiency of clinical trials. We've just published our most recent work on using machine learning to identify irregularities in randomized control trials, developed using data from 7 of PHRI's previous randomized control trials. The results are promising, suggesting that AI can improve the detection of irregularities in trials, and can do so earlier in a trial than traditional monitoring techniques. Approaches like these hold promise to help the scientific community combat the alarming trend of scientific fraud in medical research. https://lnkd.in/gxcd-4zQ

Latent class analysis (LCA) seeks to identify possible hidden classes in a sample given the observable features (descriptors, measures, etc.). In medicine it has helped identify subtypes of disease, best treatment, and pharmaceutical protein targets. We think it has future potential alongside our Glass Box Large Probability Models, e.g. to assess what classes of compounds with known different toxicities and pharmacokinetic properties to which a promising new lead compound, or group of such, might belong. But current LCA hinders that with known issues. Variables in a class are assumed statistically independent. Entrapment in local solutions is a challenge: the log-likelihood function may have a complex geography. Multiple starts can give different results. While in AI a local solution does not necessarily mean a bad solution, it may not be so when seeking clear-cut classes. There might be many or no class interpretations. Problems are worsened by sparse data, and by unknowns. Researchers inspect the classes to ensure that they are not hallucinations and balanced against that is the temptation to perceive the classifications as “truth”. Prior knowledge is important, reasonable if the aim is to show consistency with, and add further information to, an already known representation. The goal is frequently not a statistical analysis in the normal sense but to provide an interpretation easy for humans to understand, a balancing act between having just one or two classes and too many. To my taste, that seems somehow at loggerheads with LCA’s common use of entropy to better class separation and reduces human bias, but “how high” is unclear, and an overfitting can give high entropy. LCA is arguably best used in a cyclic process of validation and refinement under a critical eye, made difficult because it usually takes a great deal of computer time (hours or days) and relatedly difficulty in tackling anything remotely like Big Data. It needn't be so hard. Using information theory and number theory, with an algorithm that replaces optimization by visible and reproducible convergence to a notion of representative classes, we just completed SILCA (Systems Integrated LCA, but not a usual LCA) that runs in seconds or minutes on a standard laptop even if including further information such as diagnosis, chemical description, toxicity etc. Although the core LCA-like process is free of statistical considerations (out of respect to its very different, qualitative philosophy), the mutual information within each class is constantly checked against the many clusters generated by unsupervised high-dimensional data mining. We do not expect an exact match between two methods at opposite ends of the analytic spectrum, but significant mutual information in a class is verification of reality.   While we have a policy of publishing by peer review, there is much more real-world testing to do, and for the moment it must prove its way by serving as a promising support tool.