ARTICULAR: ARtificial inTelligence for Integrated ICT-enabled pharmaceUticaL mAnufactuRing

详细说明：人工智能用于基于 ICT 的集成制药制造

• 批准号: EP/R032858/1
• 负责人: Blair Johnston
• 金额: $ 250.4万
• 依托单位: University of Strathclyde
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2018
• 资助国家: 英国
• 起止时间: 2018 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FR032858%2F1
• 关键词: ARTICULAR; ARtificial; inTelligence; Integrated; ICT

There are considerable challenges around digitalisation in science, engineering and manufacturing in part due to the inherent complexity in the data generated and the challenges in creating useful data sets with the scale required to allow big data approaches to identify patterns, trends and useful knowledge. Whilst other sectors are now realising the power of predictive data analytics; social media platforms, online retailers and advertisers, for example; much of the pharmaceutical manufacturing R&D community struggle with modest, poorly interconnected datasets, which ultimately tend to have short useful lifespans.A result of poor, under-utilised datasets, is that it is largely impossible to avoid "starting at the beginning" for every new drug that needs to be manufactured, which is very costly with new medicines currently doubling in cost every nine years; $1 billion US Dollars currently "buys" only half a new drug so addressing this issue is key for sustainability of the industry and future medicines supply. This project, ARTICULAR, will seek to develop novel machine learning approaches, a branch of artificial intelligence research, to learn from past and present manufacturing data and create new knowledge that aids in crucial manufacturing decisions. Machine learning approaches have been successfully applied to inform aspects of drug discovery, upstream of pharmaceutical manufacturing, where large genomic and molecule screening datasets provide rich information sources for analysis and training artificial intelligences (AI). They have also shown promise in classifying and predicting outcomes from individual unit operations used in medicines manufacturing, such as crystallisation. For the first time, there is an opportunity to use AI approaches to learn from the data and models from across multiple previous development and manufacturing efforts and then address the most commonly encountered problems when manufacturing new pharmaceutical products, which are knowing: (1) the processes and operations to employ; (2) the sensors and measurements to deploy to optimally deliver the product; and (3) the potential process upsets and their future impact on the quality of the medicine manufactured.All of these data and the AI "learning" will be made available via bespoke, personalisable AR and VR interfaces incorporating gesture and voice inputs alongside more traditional approaches such as dashboards. These immersive interfaces will facilitate pharmaceutical manufacturing process design, and visualisation of the complex data being captured and analysed in real-time. Detailed, interactive 3D visualisations of drug forms, products, equipment and manufacturing processes and their associated data will be created which provide intuitive access across the length scales of transformations involved from the drug molecule to final drug product. This will be unique tool, allowing the user to see their work and engage with their data in the context of upstream and downstream processes and performance data. Virtual and Augmented Reality technologies will be used in the lab/plant environment to visualise live data streams for process equipment as the next step in digitalisation. These advanced visualisation tools will add data rich, interactive visualisation to aid researchers in their work, allowing them to focus on the meaning of results and freeing them from menial manual data curation steps.

科学、工程和制造业的数字化面临着相当大的挑战，部分原因是所生成的数据固有的复杂性，以及创建有用的数据集所面临的挑战，这些数据集具有允许大数据方法识别模式、趋势和有用知识所需的规模。虽然其他行业现在正在意识到预测数据分析的力量，例如社交媒体平台，在线零售商和广告商;大多数制药制造研发社区都在与适度的、互连性差的数据集作斗争，这些数据集最终往往具有较短的有用寿命。由于数据集较差、利用不足，对于每一种需要制造的新药，基本上不可能避免“从头开始”，这是非常昂贵的，目前新药的成本每九年翻一番;目前，10亿美元只能“购买”一半的新药，因此解决这一问题是该行业可持续发展和未来药品供应的关键。这个项目，ARTICULAR，将寻求开发新的机器学习方法，人工智能研究的分支，从过去和现在的制造数据中学习，并创造新的知识，帮助关键的制造决策。机器学习方法已成功应用于药物发现、制药上游的各个方面，其中大型基因组和分子筛选数据集为分析和训练人工智能（AI）提供了丰富的信息源。它们还显示出对药物制造中使用的单个单元操作（如结晶）的分类和预测结果的希望。这是第一次有机会使用人工智能方法从以前的多个开发和制造工作中的数据和模型中学习，然后解决制造新药品时最常见的问题，这些问题包括：（1）要采用的流程和操作;（2）要部署的传感器和测量，以最佳地交付产品;（3）要部署的传感器和测量。所有这些数据和人工智能“学习”都将通过定制的、个性化的AR和VR界面提供，这些界面将手势和语音输入与更传统的方法（如仪表板）结合在一起。这些沉浸式界面将促进制药工艺设计，以及实时捕获和分析复杂数据的可视化。将创建药物形式，产品，设备和制造过程及其相关数据的详细，交互式3D可视化，提供从药物分子到最终药物产品的整个转化长度尺度的直观访问。这将是一个独特的工具，允许用户查看他们的工作，并在上游和下游流程和性能数据的背景下使用他们的数据。虚拟现实和增强现实技术将用于实验室/工厂环境，以可视化过程设备的实时数据流，作为数字化的下一步。这些先进的可视化工具将添加数据丰富的交互式可视化，以帮助研究人员进行工作，使他们能够专注于结果的意义，并将他们从繁琐的手动数据管理步骤中解放出来。

项目成果

A Novel Integrated Workflow for Isolation Solvent Selection Using Prediction and Modeling.

• DOI: 10.1021/acs.oprd.0c00532
• 发表时间: 2021-05-21
• 期刊: Organic process research & development
• 影响因子: 3.4
• 作者: Ottoboni S;Wareham B;Vassileiou A;Robertson M;Brown CJ;Johnston B;Price CJ
• 通讯作者: Price CJ

Robust Model-Based Reinforcement Learning Control of a Batch Crystallization Process

批量结晶过程的基于模型的鲁棒强化学习控制

• DOI: 10.1109/icsc50472.2021.9666494
• 发表时间: 2021
• 作者: Benyahia B

Structural investigation and compression of a co-crystal of indomethacin and saccharin

• DOI: 10.1039/c9ce00838a
• 发表时间: 2019-08-14
• 期刊: CRYSTENGCOMM
• 影响因子: 3.1
• 作者: Connor, Lauren E.;Vassileiou, Antony D.;Oswald, Iain D. H.
• 通讯作者: Oswald, Iain D. H.

Impact of Paracetamol Impurities on Face Properties: Investigating the Surface of Single Crystals Using TOF-SIMS

扑热息痛杂质对表面性质的影响：使用 TOF-SIMS 研究单晶表面

• DOI: 10.1021/acs.cgd.7b01411
• 发表时间: 2018
• 期刊: Crystal Growth & Design
• 影响因子: 3.8
• 作者: Ottoboni S

Optimal trajectory tracking control of batch crystallization process based on reinforcement learning

基于强化学习的批量结晶过程最优轨迹跟踪控制

• DOI: 10.3390/iocc_2020-07731
• 发表时间: 2020
• 作者: Benyahia B