Information Bias in Depression

Many people will experience clinical depression in their life. While there are treatments which help, they often don't work as well or as reliably as we would like. It is therefore essential that new, more effective treatments are developed. In order to do this we must first understand the full range of interconnected mechanisms, from abnormal brain function to maladaptive habits in thinking, which cause people to become, or to remain, depressed. A better understanding of these multi-level mechanisms, and how they link together, is an essential step in the development of the new treatments which are needed for this illness.

Recently, "computational modelling" techniques have been successfully deployed to investigate the causes of psychiatric symptoms. These techniques use simple computer programs which are adjusted until they mimic the behaviour of humans and are useful at linking symptoms of illnesses to the thinking habits and brain systems which underpin them. A particular insight of computational modelling studies, compared to other nethods of understanding brain function, is that people adjust their thinking habits and brain activity to focus more on those events which provide information about the future. For example, consider a situation in which experiencing a positive event, such as being complimented, means that other positive events are more likely to occur, but in which negative events, such as being insulted, occur at random. In this situation, the positive events contain more information about the future than the negative events and because of this, as predicted by the computational modelling approach, people will tend to focus more on the positive than negative events (e.g. they will pay more attention to the positive events and learn more from them). In depression, patients focus more on negative than positive events, in other words they behave as if negative events provide more information. Indeed, this focus on negative events is believed to be one of the factors which cause people to become depressed in the first place. Computational modelling studies have linked this process of estimating the inforomation content of events with activity in the anterior cingulate cortex of the brain and of the neurotransmitter norepinephrine (NE) system suggesting that these brain systems may also be involved in producing the negative "information bias" of depression. In summary computational modelling provides a novel method for understanding the negative thinking habits which cause depression and for linking these habits to the brain systems which underpin them. In doing so it suggests new ways in which the thinking habits can be altered.

In this project I will use these insights of computational modelling to clarify the cognitive and neural mechanisms which cause depression and explore how these mechanisms may be targeted by novel treatments. Specifically, I will:

a) use computational modelling to precisely describe the information bias of depressed patients
b) use functional neuroimaging to investigate the neural mechanisms by which these information biases are acquired
c) test the degree to which the biases may be modified with a simple training interventions and with the commonly prescribed drug atomoxetine (which changes the activity of the NE system).

This translational project will be completed in the Department of Psychiatry of the University of Oxford and will include studies of healthy participants and depressed patients.

By using recent advances in computational neuroscience to better understand the causes of depression this project aims to guide the development of novel, and more effective, treatments for this devastating illness.

Depression is a common and debilitating illness. While a range of treatments are available, remission rates are low and relapse common. It is therefore essential that new, more effective treatments are developed. In order to do this we must first understand the full range of interconnected mechanisms, from abnormal brain function to maladaptive cognitive habits, which cause people to become, or to remain, depressed.

Recently, computational modelling techniques have been successfully used to link cognitive and neurobiological function with symptoms of psychiatric illness. One of the insights of this approach is that events with a high "information content", defined as the degree to which the event improves prediction of the future, are preferentially processed. Activity of the central norepinephrine (NE) system and of the anterior cingulate cortex covaries with the information content of stimuli indicating that these systems may underpin this phenomenon.

Depression is characterised by negative cognitive biases which are causally implicated in the illness. In terms of the information theoretic account described above, depressed patients behave as if negative events provided more information than positive events. In the current proposal I will use computational modelling to investigate why patients show this "information bias", what neural systems support it and how it can be altered by cognitive and pharmacological interventions.

Specifically, I will:
a) characterise the computationally defined information bias of depressed patients
b) investigate the neural mechanisms by which an information bias is acquired
c) test the degree to which it may be modified with a simple training intervention or by manipulation of the NE system using atomoxetine.

By using recent advances in computational neuroscience to better understand the causes of depression this project will guide the development of the novel treatments which are needed for the illness.

Depression is a common, chronic and disabling disorder with enormous costs to individual patients and to society. The World Health Organisation (WHO) has projected that unipolar depressive disorders will be the second leading cause of worldwide disability-adjusted life years (DALYs) by 2030. While a range of both psychological and pharmacological interventions have been proven effective in the treatment of depression, treatment response tends to be in the order of 50% (remission rates are significantly lower) with recurrence after treatment being the norm. There is therefore a strong rationale for developing interventions which treat and prevent this disorder. In order to achieve this goal it is necessary to understand the multi-level mechanisms which confer vulnerability to and maintain the illness-- and the interventions which target and modify these mechanisms. The current application, which applies recent advances in computational neuroscience in order to better understand the aetiology of depression and to guide the initial develop of novel interventions for the illness, represents the first step in this process. Ultimately therefore, the primary beneficiaries of the research are future patients. However, as described above there is also a strong economic case (both nationally and globally) for the development of more effective treatments and preventative interventions in depression.

From the point of view of industry, a key difficulty in the development of novel interventions in psychiatry has been the limited aetiological understanding of mental disorders and role of individual neurotransmiter systems in that aetiology. Such understanding is essential to reliably identify promising interventions for clinical development. The current application seeks to use recent advances in computational neuroscience to improve the aetiological understanding of depression and, more specifically, to initially test methods for altering causal processes in depression vulnerability. A better understanding of these processes increases the prospects of successfully identifying future effective treatments and therefore provides a benefit to those who seek to develop novel pharmacological and cognitive interventions.