Tuesday, 11 September 2012

Gene pathway analysis and autism

"In conclusion, we have developed an accurate diagnostic test for a genetically homogeneous group to aid in early detection of ASD". The big headline to come from this paper by Skafidas and colleagues* (open-access) reporting results based on gene pathway analysis of SNPs previously associated with autism.

This is an interesting paper for quite a few different reasons. Primarily it kinda accepts that looking at SNPs alone as being the key to autism is probably not going to be a great idea. I've talked before about the very messy science of mutation when applied to autism together with the fairly recent news that no one SNP seems to be universally related to autism, all cases of autism. Outside of the growing interest in epigenetics and autism (see here) the current paper looks beyond the SNP alone as being key; instead also looking at the cellular processes behind the SNPs and whether commonalities among ethnically homogeneous groups on the autism spectrum might yield more valuable information.

A quick summary:

  • A two-fold study primarily based on the the collected data from the AGRE initiative examining: (i) which groups of SNPs and their cellular processes might be pathogenic or protective for autism spectrum disorder(s) (ASD), and (ii) application of machine learning (artificial intelligence) to categorise SNPs which might be predictive for a diagnosis of autism. Machine learning and autism y'say? Yep.. similar to the ADOS and ADI boiling down papers recently discussed.
  • Sample numbers are a bit complicated in that both training and testing sets were used as per the method employed in the five serum metabolites and schizophrenia paper covered a while back; said testing sets derived from other databases including the SFARI database. That and the fact that ethnicity was also a factor which included a Central European cohort (CEU) and a Chinese Han cohort (HAN). We are talking in the high hundreds in terms of participant data used to derive information on which SNPs might be important.
  • Results: based on data from nearly a thousand people with an ASD in the CEU, 775 SNPs were identified as being perturbed. Thirteen different pathways were linked to these genes; six pathways crossing both the CEU and HAN. These included pathways tied into quote: "purine metabolism, calcium signaling, phosphatidylinositol signaling, gap junction, long-term potentiation and long-term depression".
  • Further analysis pertinent to predicting ASD (ASD vs. not-ASD) suggested that 237 SNPs in 146 genes were really important and "correctly predicted ASD diagnosis in 85.6% of CEU cases". When applied to the tester sets, predictive percentages fell to around 70%.
  • Eight SNPs in 3 genes (GRM5, GNAO1 and KCNMB4) were reported to be highly discriminatory in terms of the ASD or not-ASD debate; some protective and others contributory.

And rest.

To continue with the 'its interesting' theme, a few other pointers from this study are apparent. The authors for example make some discussion out of how much overlap there was in the pathways tied to the SNPs between the different ethnic groups. Certainly you would expect that even if the SNPs were in different places, similar functions might be affected which seems to be what is being reported in the current paper.

One particular pathway which cropped up caught my eye relating to purine metabolism; a source of discussion already on this blog as per this post. I don't really want to rehash that post but the watchwords are Coleman and Page, nucleotides, uric acid, uridine. Should I also at this point mention the starting material for BH4 being a nucleotide too? As for the other pathways, interesting but a little beyond my level of competence.

So the final word. Well once again, interesting. If I had to make one suggestion about improving on this work it would focus on bringing together such gene pathway analysis with other sciences such as metabolomics where both genes and functional biochemistry may very well help raise those predictive values to something very much more accurate and reliable.


* Skafidas E. et al. Predicting the diagnosis of autism spectrum disorder using gene pathway analysis. Molecular Psychiatry. September 2012


ResearchBlogging.org Skafidas E, Testa R, Zantomio D, Chana G, Everall IP, & Pantelis C (2012). Predicting the diagnosis of autism spectrum disorder using gene pathway analysis Molecular Psychiatry DOI: 10.1038/mp.2012.126