Professor Andreassen, could you describe the key objectives of the REALMENT project and how it aims to revolutionise treatment strategies for mental disorders?
The main aim of REALMENT is to bring personalized medicine interventions to psychiatry to improve outcomes for people with mental disorders. The sub-aims are to facilitate access to large amounts of Real-World Data (RWD) by developing a sustainable data infrastructure (REAL-WD platform), develop novel artificial intelligence and machine learning tools to exploit the large datasets/volumes of data and finally to combine these results to generate a clinical management platform with prediction algorithms for medication response and adverse effects in clinically helpful use cases.
Could you share the story behind the birth of the REALMENT project—what initial spark led to its conception?
The recent progress in personalized medicine approaches provides large opportunities for a range of human diseases. However, there has been little focus on developing personalized interventions in psychiatry. One of the main reasons has been the lack of large enough training data that can be used to develop prediction and stratification tools, with clinically relevant prediction accuracy. Further, there has been a lack of specific tools that can be used for psychiatric use cases. We have teamed up with partners who have access to large biobank and registry data, as well as pharma companies with clinical trial data, and machine learning expertise and formed a multidisciplinary team to move the research field forward and develop clinically useful applications in psychiatry.
Given the complexity of mental disorders and the challenge of treating affective and psychotic symptoms, how does the REALMENT project plan to optimise the use of existing pharmaceutical treatments?
The main challenge is the large variation in treatment response, with nearly one third being non-responders to first line treatments. However, we are currently using a trial and error approach in the clinic, and patients may receive inefficient treatments with severe side effects for long periods. Recent findings have shown that treatment response have a heritable component, and we will use genetic profiles to optimise the use of existing medications. In this way, we can reduce side effects and increase the effects of the treatment for patients with mental disorders. To bring about such precision psychiatry, we will make use of large data sets from electronic health records and registers, as well as new methods such as machine learning and artificial intelligence
Given the significant clinical problem of non-response to medication and large variation in efficacy and adverse effects, how will REALMENT's approach of optimising the use of existing medication based on a patient’s genetic profile address these issues?
It has become increasingly clear that mental disorders are caused by multiple factors (genetic, environment, brain function, psychosocial stressors), that work together in disease development. While there is a large genetic susceptibility component, there is not a single gene that lead to a mental disorders, but many hundreds of variants that act together, each with a small effect. Similarly, response to medication is determined by multiple small effects. Thus, in order to identify the individual risk factors, we need to train prediction tools on several thousand samples, which can only be achieved with real world data. Then we will apply the genetic profile of a given patient, to help prediction effect and side effect of a given drug, and translate this into a tool to guide the patient and the doctor in choosing the best treatment for the individual, balancing potential effect and side effect risk.
Could you explain how the REALMENT project plans to use AI and machine learning to improve treatment outcomes and patient quality of life?
We will use Artificial intelligence tools, such as supervised machine learning, to exploit the massive amount of data. For example, we have treatment outcome from thousands of patients which we will use to generate genetic profiles that are associated with response or side effects, and quality of life. The tools will be fed with prior knowledge (about pharmacological profile, know side effects etc), and then analyse the data for patterns of data that can be used to predict for example side effects of lithium or antipsychotic medication. The response prediction will not be 100% accurate, but will provide some added value that the clinician can add to the current guidelines for selecting medication.
Is there something else you think the REALMENT audience should know about the project? Especially mental health patients.
It is important to take into account new opportunities that become available in medicine, and make sure they are applied across all fields and specialties. Until now, most focus and funding in precision medicine have been in cancer research. Still, there is a large potential in mental illness, which can form the basis of better treatment. It is also important that there are some challenges withing precision medicine, such as data security, expertise of clinicians, and the ability for technical solutions that can be implemented in real world clinics. Further, there are some ethical challenges with precision medicine, and it is important for user organisations and people with lived experiences to know what precions medicine can provide of new opportunities, but also challenges. Thus, we have in our project a strong Stakeholder forum and we are discussions of ethical challenges are integrated in the research process.