Five serum metabolites and schizophrenia

This post extends my fascination with all things metabolomics, biomarkers and behaviourally-defined conditions. This is quite an apt post for me at the moment given that the all-singing, all-dancing Q-ToF mass spectrometer we have at work is currently receiving a little TLC from the appropriate engineer after a few days being under the weather (P.S. get well soon xxx).

I have previously talked about research presented on schizophrenia and the hunt for suitable biomarkers on this blog. Remembering back to the casein antibody link and the more recent offering on amino acids, a new study has just emerged by Yang and colleagues* with the rather interesting suggestion that just five serum markers (and one urinary marker) might be able to separate schizophrenia from not-schizophrenia equivalent to diagnostic interview.... read on.

The paper which is open-access is pretty comprehensive (aren't they always) but I will try and summarise the methods and main findings here:

• Serum and urine samples from Chinese participants diagnosed with schizophrenia (various onset psychosis) were compared with asymptomatic controls. The absence of diabetes, heavy alcohol consumption and being two weeks mood-stabilising drug free were prerequisites for study entry.
• Enrolled participants were allocated to a training set (N=124) or a test set (N=98). Sample combinations were analysed by gas chromatography - Time of Flight mass spectrometry (GC-ToF) and NMR. These methods and their important accompanying software, represent gold-standards in metabolomic analysis; the ToF bit gives you accurate mass of your compounds (mass-to-charge ratio of an ion) and the NMR, structural information about your compounds of interest.
• Principal component analysis (PCA) provided some information on how the various compounds related to the groups, and after applying a filter on the data, the detected compounds were whittled down to those which gave most discriminating power between schizophrenia and not-schizophrenia. 
• In the end five compounds detected in serum were deemed most important to discriminating schizophrenia in this participant group: glycerate, eicosenoic acid, beta-hydroxybutyrate, pyruvate and cystine. With this compound set, the receiver operating characteristics (AUC) were 0.945 in the training samples and 0.895 for the test samples. When a urinary compound was added to the set (beta-hydroxybutyrate), the AUC went to 1 (which represents a perfect classification) in both training and test sets.

OK you can perhaps see why this paper and its data is so exciting. Move over diagnostic interview, hello serum and urine analysis? The authors delve a little deeper into the compounds reported in their findings and their potential relationship to schizophrenia. At least three of the compounds recorded are tied into the degradation of glyceride and fatty acids which might point to energy metabolism as being implicated in schizophrenia. Some of this might (might!) also link back to related findings in certain types of autism and in particular talk of mitochondrial issues? Cystine is also discussed a little more in the paper and again, some familiar words to autism in relation to things like glutathione and homocysteine; I also wonder whether the findings of low cystine in urine and high in serum might also tie into some of the sulphation issues so unfortunately lost from the current autism research landscape.

As always there are some important things to say about this study before we get too carried away. First is the population used; all Chinese. Whether such markers are applicable to populations outside of China remains to be seen. Second is the relatively small participant group looked at and issues relating to any comorbidity which have not really been fully addressed. Finally are the comparisons of the findings with other, similar studies. The Rujescu findings mentioned in my last post on amino acids suggested that four amino acids might be discriminatory for schizophrenia. As far as I can tell, none of these amino acids, in either plasma or urine, showed up on the current paper (allowing for point one on different peoples and different profiles).

I will perhaps read and re-read the current paper just to make sure that I am seeing all that is being said. Likewise I might just flick through some of the autism research in this area (such as the Yap paper from last year) just to see if there are any other areas of 'overlap'.

* Yang J. et al. Potential metabolite markers of schizophrenia. Molecular Psychiatry. October 2011.