This publish was written with Avinash Erupaka from Bayer (IT PH, Drug Innovation platform)
How can pharmaceutical firms unlock the complete potential of their information to drive breakthrough improvements? Bayer, a world chief in well being and diet, is devoted to tackling the urgent challenges of our time, together with a rising and ageing inhabitants and the pressure on our planet’s ecosystems. Its mission of “Well being for All, Starvation for None” drives its dedication to addressing societal and environmental wants by way of groundbreaking analysis. Bayer is targeted on growing modern options that make a tangible distinction on the earth and worth for its prospects, staff, and stakeholders. Headquartered in Leverkusen, Germany, Bayer operates throughout 80 international locations and is pioneering a knowledge science ecosystem that transforms how analysis groups entry, analyze, and derive insights from complicated scientific information.
By harnessing the facility of knowledge, analytics, synthetic intelligence and machine studying (AI/ML), and generative AI, Bayer is making a cloud-based Pharma R&D Knowledge Science Ecosystem (DSE) on AWS that powers cutting-edge applied sciences and ideas with sturdy information administration. In doing so, R&D groups can totally understand the potential of unified information and analytics.
On this publish, we focus on how Bayer used the subsequent era of SageMaker to construct an answer that unified information ingestion, storage, analytics, and AI/ML workflows. Constructed on information mesh rules, Bayer’s DSE integrates superior information ingestion, storage, analytics, and ML workflows to allow agile experimentation and scalable perception era. It democratizes entry to analytics, fosters cross-Area collaboration, and supplies versatile integration of structured, semi-structured, and unstructured information.
Challenges in pharmaceutical analysis
In pharmaceutical analysis, information has develop into essentially the most crucial asset for driving innovation. Nevertheless, managing this information successfully presents unprecedented challenges and conventional information administration approaches have gotten more and more insufficient for complicated, world analysis initiatives. Many pharma R&D group face a fancy ecosystem of knowledge and analytics associated obstacles that hinder scientific discovery and operational effectivity:
- Siloed datasets – Analysis datasets are siloed throughout domains, limiting reuse and slowing discovery.
- A number of information modalities – Medical trial information (structured), real-world proof (semi-structured), and genomic information (unstructured) existed in isolation, complicating integration and evaluation.
- Rigid ingestion capabilities – Techniques that assist batch processing (akin to trial information), real-time information streams (for instance, from lab tools), and event-driven ingestion (akin to regulatory updates).
- Rising R&D prices – Disparate applied sciences and disconnected programs create operational inefficiencies and elevated licensing and upkeep prices.
- Inconsistent panorama to totally use ML – The absence of a unified information structure and standardized, domain-agnostic MLOps workflows imply that information and analytics innovation is usually advert hoc and non-repeatable. Groups lack a streamlined solution to scale profitable patterns, leading to redundant efforts, longer improvement cycles, and missed alternatives for cross-domain synergy.
- Disconnected architectures – Software program options should not built-in into the broader unified ecosystem, leading to silos, redundancies, and inefficiencies.
Recognizing these systemic challenges, Bayer launched into a transformative journey. DSE is not only a technological resolution, however a strategic reimagining of how analysis information and analytics may very well be used throughout a world group. By bringing collectively cutting-edge applied sciences, standardized frameworks, a collaborative information mesh, and lakehouse structure, Bayer got down to assist researchers and engineers speed up pharmaceutical innovation.
Discovering an answer with the subsequent era of SageMaker
Bayer envisioned a unified information science ecosystem that would supply the next:
- A unified collaborative improvement expertise for all information scientists no matter their location or specialization
- Seamless entry to each structured and unstructured information by way of a constant interface
- Constructed-in governance and compliance controls applicable for pharmaceutical analysis
- Scalable compute assets to deal with essentially the most complicated analytical workloads
Bayer performed a complete analysis of varied options earlier than deciding on the subsequent era of SageMaker because the cornerstone of their new information science ecosystem. Though different choices had deserves, Bayer prioritized the next capabilities:
- Entry to multimodal information – Important for genomics, proteomics, and superior biomarker analysis
- Centralized asset market – Central hub to find and reuse information, options, fashions, and different enterprise belongings
- Built-in tooling ecosystem – Streamlined entry to key instruments like Git, ETL, MLflow, and generative AI utility builders in a single place
- Multi-domain and cross-Area assist – Vital for world analysis collaboration
- Value-performance – Needed for sustainable, long-term scaling
The capabilities of Amazon SageMaker Unified Studio and Amazon SageMaker Catalog aligned with Bayer’s imaginative and prescient of decentralized mesh execution mixed with centralized discovery and governance. They enabled groups to work with their most well-liked instruments, akin to Jupyter Notebooks or workflow builders, whereas sustaining discoverability and reusability of belongings.
Answer overview
This part describes the important thing options and structure of Bayer’s DSE constructed on SageMaker. The DSE resolution addresses the recognized challenges by way of a multi-layered structure:
- Breaking down information silos – Multimodal information ingestion capabilities of the answer break down information silos by enabling unified storage, processing of structured, semi-structured, and unstructured information by way of batch, streaming, and event-driven pipelines.
- Dealing with various information modalities – A hybrid lakehouse structure, constructed on Amazon Easy Storage Service (Amazon S3), Apache Iceberg, and Amazon Redshift, supplies a versatile basis for dealing with various information modalities and maturities whereas offering information consistency and accessibility.
- Decreasing prices by way of standardization – To deal with rising R&D prices and operational inefficiencies, pre-wired analytical workbenches supply standardized templates and built-in improvement environments (IDEs) that scale back redundancy and speed up workflow improvement.
- Unlocking AI/ML with Amazon SageMaker AI and Amazon Bedrock – Superior AI/ML capabilities, powered by Amazon SageMaker AI and Amazon Bedrock, create a standardized, domain-agnostic MLOps surroundings that allows repeatable innovation and cross-domain synergy.
- Managing instruments ecosystem with end-to-end observability – Strong governance and observability options present compliance and system reliability whereas integrating beforehand disconnected instruments right into a unified, well-monitored ecosystem that breaks down architectural silos and promotes environment friendly useful resource utilization.
The DSE structure implements information mesh rules the place information domains (omics, regulatory, medical trials) are handled as merchandise, with possession and administration obligations assigned to area specialists. These domains are decentralized for execution however stay discoverable and reusable by way of SageMaker Catalog. On the core of the structure is a hybrid mesh lakehouse structure that mixes Amazon S3 and Iceberg, offering the flexibleness to deal with each structured and unstructured information effectively. SageMaker Unified Studio supplies an analytical layer the place researchers can entry the complete suite of instruments wanted for his or her work. The next diagram illustrates this structure.
Influence
The primary section of Bayer’s DSE confirmed the subsequent era of SageMaker as a robust basis for his or her R&D DSE—designed to steadiness decentralized innovation with centralized governance by way of a scalable information mesh structure. With this resolution, Bayer can catalog and handle multimodal information belongings—together with structured and unstructured information, ML options, fashions, and customized scientific belongings—with context-rich metadata throughout various Pharma R&D domains. Bayer is now positioned to onboard over 300 TB of biomarker information and combine siloed omics, medical, and chemistry information repositories right into a cohesive surroundings. With built-in instruments like JupyterLab Areas, MLflow, and SageMaker AI Studio, the DSE platform is laying the groundwork for a complete, GxP-aware ML workbench—paving the best way to operationalize over 25 high-value ML use instances and assist greater than 100 information scientists throughout the group.
“The Knowledge Science Ecosystem is important for growing our medicines,” says Daniel Gusenleitner, Mission Lead for the R&D Knowledge Science Ecosystem. “It enhances our enterprise workflows with superior analytics, serving to us speed up the seek for new therapies. By integrating information from all the analysis and improvement course of, we enhance the probabilities of technical success and guarantee our efforts are environment friendly. Unlocking our information additionally facilitates goal discovery, resulting in groundbreaking developments in affected person care.”
Subsequent steps
Bayer has efficiently begun their Knowledge Science Ecosystem on the subsequent era of Amazon SageMaker and is working to onboard the primary use case of superior biomarker analysis. Constructing on the sturdy basis, Bayer can be accelerating the evolution of the DSE resolution with the next key enhancements:
- Federated catalogs and cross-domain integration – Enabling search and reuse of knowledge belongings throughout therapeutic areas and enterprise items
- Superior ontology and semantic layer – Enriching metadata with area information to assist AI-based search, discovery, and reasoning
- Adoption of generative and agentic AI workflows – Driving novel drug discovery and accelerating speculation era
Conclusion
By leveraging the subsequent era of Amazon SageMaker to construct their cloud-based Knowledge Science Ecosystem, Bayer is making a basis for quicker, extra environment friendly analysis and discovery. Amazon SageMaker is unifying various information varieties, enabling world collaboration, and standardizing ML workflows to assist place Bayer on the forefront of data-driven innovation.
To study extra and get began with the subsequent era of SageMaker, check with Amazon SageMaker or the AWS console.
Concerning the Authors
