The results from our data indicate that three distinct neurophysiological profiles are evident in our patient population. The largest subgroup that was identified is characterised by a global impairment of the MMN/P3a complex, as reflected by the reductions in frontal and temporal MMN and P3a amplitudes. The +TMMN cluster bears similarities to the globally impaired cluster in terms of deficits in frontal MMN amplitudes, however was unique from the other two clusters by exhibiting the largest temporal MMN amplitudes. Finally, the +FMMN cluster was distinguished from the other clusters in that it had the largest frontal MMN amplitudes. Notably, no differences were observed between all three cluster-groups for any clinical and cognitive measures except that the globally impaired cluster displayed poorer verbal memory compared to the +FMMN cluster.
Just over half (55%) of the patients in the globally impaired cluster comprised of individuals with a primary diagnosis of a psychotic disorder. Furthermore, the global impairments identified in this cluster are consistent with the extant literature which reports that deficits in frontal and temporal MMN amplitudes are well established findings in schizophrenia, and more recently, earlier stages of psychotic disorders 
. Moreover, it is well established that these neurophysiological components reflect the integrity of the neural pathways that they represent and as such, the relative health of these pathways can be gleaned via measures such as amplitude as well as from the corresponding morphology of these signals 
. In the case of the globally impaired cluster, the observed decrement in amplitudes for both temporal (MMN) and frontal (MMN) signals suggests that for these patients there may be a global disruption in the MMN subsumed by deficits in both the sensory memory and the automatic attention-switching mechanisms underlying deviance detection 
. Furthermore, this neurophysiological profile corresponded with the worst overall cognitive profile; a finding that is consistent with significant associations reported between MMN and a range of cognitive variables 
. This association should be treated with caution as although the globally impaired cluster was the only group to show significant reduction across all neuropsychological measures compared to controls, there was only one significant difference among the cluster groups (see pair-wise comparisons in ). Potentially, such functional associations may be long-term, with other evidence showing that patients with the greatest MMN impairments at baseline had the worst functional outcomes at follow-up 
These findings indicate that a subgroup of patients, at the early stages of a psychotic disorder, already display significant neurobiological impairment. In other words, a global reduction in MMN/P3a with corresponding impairments in cognitive performance may be a biomarker for a more distinct and/or prolonged psychotic (or related) illness. However, it is important to keep in mind that 46% of the globally impaired cluster consists of individuals with a primary affective disorder (i.e. a depressive or a bipolar disorder with and without psychotic features). It is possible that, especially given their younger age, that some of these patients may be on a pathway to a psychotic illness (with the depressive phenotype currently dominating). Notwithstanding, impaired MMN and P3a has been documented in individuals with affective disorders 
and a recent study by our group has reported impaired temporal MMN in an older sample of patients with late-life depression indicating a specific disruption in temporal networks 
. Similarly, in psychotic disorders, significant impairments in temporal MMN tend to be found in chronic samples 
. Critically, no studies have undertaken a data-driven approach to investigate the extent to which there may be subsets of patients with broad (in particular, temporal) impairments in MMN so it is difficult to determine whether global impairments in MMN/P3a represent severity, chronicity or both. In all likelihood, the neurophysiological profile shown by the globally impaired cluster may not be related to a particular diagnosis or syndrome but rather represent more severe deficits that are indicative of a developmental origin and related to general brain dysfunction. As such, this may be a larger risk factor for psychotic (and related) disorders, but not exclusively.
The +TMMN cluster bears similarities to the globally impaired cluster in that they both show deficits in frontal MMN. Frontal MMN impairments have been reported in both chronic schizophrenia and also early stages of psychotic illnesses, such as first-episode psychosis and ultra-high risk for psychosis 
. On the other hand, only one study has reported impairments in temporal MMN at early stages of psychotic illnesses 
; a finding which is more common in chronic schizophrenia samples 
. We have proposed that deficits in temporal MMN may develop with severity and/or chronicity 
. In light of this, the absence of impairment in temporal MMN in the +TMMN cluster may indicate that individuals are less severe or at an earlier stage of their illness compared to those in the globally impaired cluster. The +TMMN cluster is equally represented by individuals with psychotic (41%), bipolar (41%) and depressive disorders (18%). Notably, Hafner and colleagues determined that the most common and stable symptoms across affective and psychotic disorders are depressive symptoms, negative symptoms and functional decline 
. This research suggests that at early stages of both psychotic and affective disorders, the longitudinal trajectory is often unclear. Therefore, we speculate that of the individuals diagnosed with affective disorders in the +TMMN cluster (who may not currently display positive symptoms of psychosis), some might display such symptoms in the future and show more widespread neurophysiological deficits indicating further neurobiological changes.
The most surprising finding in this study was the significantly increased frontal MMN which characterised the +FMMN cluster. Increased frontal MMN is thought to reflect a hyper-glutamatergic state such as that seen in those who are prone to alcoholism 
. Of note, this cluster had the highest, albeit non-significant, AUDIT total scores (see ) suggesting some association with risky drinking. These individuals may represent a distinctly separate phenotype, despite symptomatology or diagnostic category, and may have quite different treatment implications, particularly given the recent interest in treating psychotic illnesses with glutamatergic agents 
. NMDA receptor hypo-function and dopaminergic hyper-function are a widely accepted phenomenon of psychotic disorders 
. These neurotransmitter systems are closely linked since glutamate activates the inhibitory systems of dopamine and this regulatory response is necessary to maintain equilibrium of these neurotransmitters 
. Antipsychotics are primarily dopamine antagonists which reduce dopamine levels but also increase glutamate levels 
. Excessive activity of NMDA receptors has been implicated in mood disorders 
and treatment with mood stabilisers (i.e. lithium and lamotrigine) or antidepressants has been shown to reduce the glutamate levels associated with such activity 
. A better understanding of the MMN profiles revealed in this study may help to determine whether different types of glutamatergic agents (if any) may be efficacious in treating certain individuals.
Although this is the first study using a data driven approach to determine MMN/P3a complex profiles in young people with psychiatric disorders, it is limited by its cross-sectional nature. For future studies, data at multiple time points across illness trajectory will help to better determine the temporal nature of illness. Additionally, by better defining stage and severity of illness the predictive capacity of these cluster profiles can be assessed. The cohort in this study was a sample of convenience and overall there are unequal ratios of primary psychotic, bipolar and depressive disorders; future studies should ideally have larger sample sizes with equal ratios of diagnostic category. Lastly, this analysis is exploratory and further research is warranted to replicate these findings.
This study highlights the utility of using a data driven method as an alternative to categorising individuals with psychiatric illnesses. In this respect, this study supports the notion that MMN and P3a are well suited to probe the underlying neurobiology of psychiatric disorders and provide important insights into the variations of neurochemical functions that appear to exist despite diagnostic category.