DX in Research
The process of research involves obtaining a hit compound (a compound that can readily bind to a disease-causing target molecule) and optimizing it into a drug candidate compound (a compound with enhanced suitability as a drug). This used to be a costly and time-consuming process, but Astellas is working to improve it through the implementation of DX.
Ultra Large scale Virtual Screening
By performing high-speed calculations and assessing a larger number of compounds, we can obtain potent hit compounds in a short period of time. In the past, calculations were performed on approximately one million compounds using an internal server. It would have been possible to obtain more potent hit compounds if we were to evaluate hundreds of millions of compounds. However, the process would have required one to two years on the conventional system. We have updated our system by combining cloud computing with AI that predicts ease of binding. Now, we can complete calculations for hundreds of millions of compounds in as short as one to two weeks.
“Human-in-the-Loop” Drug Discovery Platform
Furthermore, we are also working to shorten the research time to optimize acquired hit compounds into high-quality drug candidates. We installed AI and robots to accelerates revolving a DMTA cycle where we "Design" and "Make" the compound, "Test" the effects, "Analyze" the information, and design better compounds for the next cycle based on the results obtained from the previous cycle. The speed of drug discovery has been dramatically improved by integrating AI and robots into each step, with researchers adding value in the form of ideas and comprehensive judgment at key stages. Using this platform, we have reduced the time it takes from hit compound to acquisition of a drug candidate compound by approximately 70% in successful cases.
In order to utilize this platform for new modalities such as cells and genes, the robot Mahol-A-Ba is being used in the Test phase to evaluate pharmacological effects.
Astellas DX Strategy Series Vol.2: Drug Discovery Platform Integrating Humans, AI, and RobotsRead more
E-PaD: A Research Support Tool that Uses NLP
E-PaD helps provide insights necessary for drug discovery research and assists in generating new ideas for drug discovery. It uses an AI technology called NLP (Natural Language Processing) to efficiently extract meaningful information on pathology, genes, biology, modalities, etc., from the life science literature, which grows by more than 1.6 million every year. The tool not only performs simple keyword searches but also suggests highly relevant disease (indication) candidates with rationales to researchers based on the content of sentences and phrases related to biology, etc. E-PaD also has a function to consolidate and visualize information on the external environment, and can support planning of research strategy based on competitive landscape and life science trends.
DX in Development
Hybrid clinical trials
Decentralized Clinical Trials (DCT) is an initiative to reduce the burden on patients and clinical trial sites by introducing and utilizing technologies such as telemedicine and wearable devices. In order for DCT to become a permanent fixture on the clinical trial landscape, it is imperative to balance the convenience of the technology with patients’ individual needs. Astellas aims to design hybrid clinical trials that combine technology with voices from patients.
The initial phase will include “piloting with a subset of patients on an existing trial who would take two assessments instead of just the one—the traditional assessment and then the technology-enabled assessment.” Gathering data this way is key to building the proof of concept necessary to gather momentum and support for the hybrid model. An example of how this could be put into practice is an oncology trial utilizing a wearable and personalized data-collection app. With the appropriate infrastructure in place, the data collected could inform a treating physician of potential medical issues, provide Astellas more robust information of the potential side effects of the product and allow the patient to feel safe in an environment more friendly than a hospital.
A New Approach to Clinical Trials Powered by Patient Voice and TechnologyRead more
DX in Manufacturing
DAIMON—An Original Data Mining System Pharmaceutical Production
The pharmaceutical industry handles a vast amount of data, not only in research and development but also in production. Producing Active Pharmaceutical Ingredients (APIs) involves chemical reaction with structural change and purification. Furthermore, the manufacturing process of products using APIs includes lots of processes such as uniformly mixing APIs, tableting, all of the processes are strictly implemented and controlled based on data.
These highly rigorous production processes are necessary for Astellas to deliver on our commitment to provide a stable supply of high-quality pharmaceuticals to patients. And as we endeavor to maintain and improve those processes, Astellas independently developed DAIMON, a state-of-the-art data mining system.
The DAIMON data mining system for manufacturing incorporates (1) univariate monitoring, (2) cause and effect and regression monitoring, and (3) multivariate monitoring. Significant time reductions in data analysis can be achieved through the appropriate operation of these three monitoring types, according to the volume of data and the complexity of the analysis.
DAIMON's strength lies in a continuous cycle of knowledge acquisition. Starting with monitoring trends, the cycle then detects fluctuations/predicts risks, investigates causes/prevents problems and improves understanding of products and processes. Through this cycle, it is possible to promptly react to quality or production trouble and prevent such troubles in advance, thereby realizing a high manufacturing level and a stable supply of pharmaceutical products.
DX in Commercial & Marketing
Utilizing Real-World Data
Gaining an enhanced understanding of patient and physician dynamics allows the company to better formulate strategies. Real-world data is used in order to prepare the commercial organization for a product launch in a new therapeutic area.
For example, longitudinal patient data is used to estimate patient population sizes and segment patients to inform reimbursement strategies and understand resource requirements. It is also used to investigate scenarios for life cycle management, including new indications and formulations.
Time series and machine learning analyses are also used to examine patient journeys and to estimate market potential. Natural language processing and social media listening are leveraged to identify the language patients use and the burdens they face in order to inform patient education materials.
DX in Corporate Functions
We are utilizing data analytics not only in our value chain but also for important strategic decision-making through corporate functions.
Using inductive data-based statistical science techniques and deductive mathematical modeling and simulation techniques, we will quantitatively analyze future uncertainty to improve the speed and quality of various strategic decisions.
We are digitizing real-time information updates on our internal pipeline and external conditions. Our portfolio analysis and quantitative risk analysis techniques, which utilize simulation technology, allow us to analyze and evaluate a wide variety of scenarios and contribute to the construction of ideal portfolios from both short- and medium- to long-term perspectives.
We are introducing analytics to evaluate business value and aim to improve the quality of our decision-making by implementing an innovation evaluation. We utilize corporate finance theory and statistical simulation techniques to quantitatively predict and evaluate the uncertainty and future potential of business value.
By utilizing time series analysis of sales and various costs and machine learning/AI-based forecasting models, we aim to maximize corporate value by identifying opportunities to improve profitability, reduce costs, and achieve the most efficient resource allocation.
We are taking on the challenge of talent management digitalization and employee network analysis by utilizing HR data-based analytics, predictive modeling, and natural language processing alongside other technologies. We conduct influencer and organizational analysis not only through inductive data analysis but also through deductive approaches utilizing behavioral economics and agent-based models.