Application of artificial intelligence to predict biologic systemic therapy clinical response, effectiveness and adverse events in psoriasis

该项目旨在改善严重牛皮癣的治疗方法。牛皮癣是一种常见的疾病，会导致红色鳞状皮肤斑块。它会造成身体、社会和心理上的痛苦，并可能导致其他长期疾病的发展。有一组强大而有效的药物，称为生物制剂，用于治疗严重的牛皮癣。然而，这些药物价格昂贵，而且可能有副作用。目前，还没有办法知道哪种药物对特定患者最安全有效。这导致医生采取“试错”的方法。这种方法可能导致对最初开出的药物反应不佳，从而导致疾病控制的延误。因此，医生需要能够在一开始就确定哪种药物最有可能改善患者的牛皮癣，平衡潜在的副作用风险。这种进步可以使疾病在早期阶段得到显著改善，从而减少患者的痛苦并减少NHS的支出。这种个性化的治疗方法已被英国牛皮癣协会确定为十大研究重点之一。被称为“人工智能”的计算机技术的新发展可以通过告知哪种治疗方法对特定患者最有效来帮助解决这个问题。某些人工智能模型可以做出比人类大脑更好的决策。与患者一起工作，皮肤科医生可以简单地将相关的患者因素输入到这样的模型中，以确定最有效的个性化治疗。创建这些模型需要大量的患者数据集，例如英国皮肤科医生生物制剂和免疫调节剂协会注册（BADBIR），这是一个世界领先的英国牛皮癣患者数据库。这些人工智能模型尚未应用于大型“现实世界”牛皮癣数据集。我已经学会了如何编写计算机代码，并获得了访问这个BADBIR数据集的权限。这使我能够创建两个初步的人工智能模型来预测对生物药物的反应。该项目的目的是提高我们对这些药物反应变化的理解，并使开发一种支持临床实践决策的工具成为可能。这将更好地指导个体患者的药物选择。该项目的总体影响将是患者护理和结果的显着改善，同时也减少了因无效治疗而浪费的NHS预算。使用BADBIR数据，该项目的目标是：1。预测用于治疗牛皮癣的生物制剂的有效性和副作用的发展。看看病人的哪些因素有助于做出这些预测。观察病人对生物药物的不同反应，并进一步了解这方面的情况，包括病人服用药物顺序的相关性。为了重复临床试验，使用数据来比较未作为一线使用的生物反应，我将探索不同人工智能模型的应用。我将使用方法来确保我们理解这些模型的内部工作原理。德国牛皮癣患者登记，PsoBest，将被用来证明我的发现可以推广到其他数据集。一项针对纽卡斯尔牛皮癣患者的调查显示了对该研究的积极支持。从这些患者中，我成立了一个专门的PPI小组，他们帮助制定了拟议的研究，并将继续指导该项目。我的项目之所以重要，有几个原因。牛皮癣是一种常见的使人衰弱的疾病，这个项目有可能改变皮肤科医生治疗这种疾病的方式。如果将我的方法应用于一系列其他专业的治疗，我希望它能产生更广泛的影响。

This project aims to improve the way in which severe psoriasis is treated. Psoriasis is a common condition that causes red, scaly skin plaques. It causes physical, social, and psychological suffering and can lead to the development of other long-term diseases. There are a group of powerful and effective drugs, called biologics that are used to treat severe psoriasis. These drugs however, are costly and can be associated with side effects. Currently, there is no way of knowing which drug will work in the most safe and effective way for a particular patient. This leads to doctors adopting a "trial and error" approach. This approach can result in a poor response to the drug initially prescribed, leading to delays in disease control. There is therefore a need for doctors to be able to identify at the outset, which drug will most likely improve a patient's psoriasis, balanced against the potential risk of side effects. Such an advance could enable dramatic disease improvement at an earlier stage, thereby reducing patient suffering and decreasing NHS spending. This personalised treatment approach has been identified by the Psoriasis Association UK patient body as one of the top-10 research priorities.New developments in computer technology, called 'artificial intelligence' could be used to help resolve this issue, by informing which treatment will work best for a particular patient. Certain artificial intelligence models can make better decisions than our human brain is capable of. Working with the patient, dermatologists could simply input relevant patient factors into such a model to identify the most effective personalised treatment. Creating these models requires large patient datasets, such as the British Association of Dermatologists Biologics and Immunomodulators register (BADBIR), a world-leading UK psoriasis patient database. These artificial intelligence models have not yet been applied to large 'real-world' psoriasis datasets. I have learnt how to write computer code and gained access to this BADBIR dataset. This has enabled me to create two preliminary artificial intelligence models to predict response to biologic drugs.The aim of this project is to improve our understanding of the variation we see in response to these drugs and enable the development of a tool that supports decision-making in clinical practice. This will better guide the choice of drug for individual patients. The overall impact of the project will be a significant improvement in patient care and outcomes, whilst also reducing wasteful use of the NHS budget through ineffective treatments.Using BADBIR data, the objectives of the project are:1. To predict a) the effectiveness and b) the development of side effects of biologics used to treat psoriasis 2. To look at which patient factors contribute to being able to make these predictions 3. To see how patients respond differently to biologic medications and understand more about this, including the relevance of the sequence of drugs a patient is prescribed4. To replicate a clinical trial using data to compare biologic response when not used as first-lineI will be exploring the application of different artificial intelligence models. I will use methods to ensure that we understand the inner workings of these models. The German psoriasis patient registry, PsoBest, will be used to show that my findings can be generalised to other datasets.A survey of Newcastle psoriasis patients showed positive support of this research. From these patients I formed a dedicated PPI group, who have helped shape the proposed research and will continue to guide the project. My project is important for several reasons. Psoriasis is a common and debilitating condition, and this project has the potential to transform the way that dermatologists manage this disease. I expect my methodology to have a much broader impact, if applied to treatments across a range of other specialities.