Phoenix Rising tells QMUL: release the PACE trial data
Mark Berry, Acting CEO of Phoenix Rising, presents the Board of Directors’ open letter to Queen Mary University of London (QMUL) urging them to release the PACE trial data, and hopes that other non-UK organisations will join British charities in the same request...
Discuss the article on the Forums.

New Jonathan Kerr gene expression paper in J Clin Pathol

Discussion in 'Latest ME/CFS Research' started by Daisymay, Sep 20, 2014.

  1. Daisymay

    Daisymay Senior Member

    Messages:
    716
    Likes:
    3,874
    http://jcp.bmj.com/content/early/2014/09/19/jclinpath-2014-202597.short?rss=1

    J Clin Pathol doi:10.1136/jclinpath-2014-202597
    • Original article
    Use of single-nucleotide polymorphisms (SNPs) to distinguish gene expression subtypes of chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME)
    1. Nana Shimosako1,
    2. Jonathan R Kerr1,2
    +Author Affiliations

    1. jonathan@ssl-mail.com
    • Received 30 July 2014
    • Revised 2 September 2014
    • Accepted 4 September 2014
    • Published Online First 19 September 2014
    Abstract
    Aims We have reported gene expression changes in patients with chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME) and the fact that such gene expression data can be used to identify subtypes of CFS/ME with distinct clinical phenotypes. Due to the difficulties in using a comparative gene expression method as an aid to CFS/ME disease and subtype-specific diagnosis, we have attempted to develop such a method based on single-nucleotide polymorphism (SNP) analysis.

    Methods To identify SNP allele associations with CFS/ME and CFS/ME subtypes, we tested genomic DNA of patients with CFS/ME (n=108), patients with endogenous depression (n=17) and normal blood donors (n=68) for 504 human SNP alleles located within 88 CFS-associated human genes using the SNP Genotyping GoldenGate Assay (Illumina, San Diego, California, USA). 360 ancestry informative markers (AIM) were also examined.

    Results 21 SNPs were significantly associated with CFS/ME compared with depression and normal groups. 148 SNP alleles had a significant association with one or more CFS/ME subtypes. For each subtype, associated SNPs tended to be grouped together within particular genes. AIM SNPs indicated that 4 subjects were of Asian origin while the remainder were Caucasian. Hierarchical clustering of AIM data revealed the relatedness between 2 couples of patients with CFS only and confirmed the overall heterogeneity of all subjects.

    Conclusions This study provides evidence that human SNPs located within CFS/ME associated genes are associated with particular genomic subtypes of CFS/ME. Further work is required to develop this into a clinically useful subtype-specific diagnostic test.
     
    Last edited: Sep 20, 2014
    taniaaust1, August59, cigana and 10 others like this.
  2. Daisymay

    Daisymay Senior Member

    Messages:
    716
    Likes:
    3,874
    The paper above seems to be this "completed study" funded by ME Research UK back in 2010, which I presume wasn't published at the time?


    http://www.meresearch.org.uk/our-research/completed-studies/single-nucleotide-polymorphisms/

    Authors
    Shimosako N, Kerr JR

    Institution
    George’s University of London, London UK

    Abstract
    We have recently reported gene expression changes in patients with Chronic Fatigue Syndrome/Myalgic Encephalomyelitis and the utility of gene expression data to identify subtypes of CFS/ME with distinct clinical phenotypes (J Infect Dis, 2008). Due to the difficulties in using a comparative gene expression method as an aid to CFS/ME disease and subtype-specific diagnosis, we attempted to achieve such a method based on single nucleotide polymorphisms (SNP) alleles.

    To identify SNP allele associations with CFS/ME and CFS/ME subtypes, we tested genomic DNA of CFS/ME patients (n=108), endogenous depression patients (n=17), 13 and normal blood donors (n=68) for 454–504 human SNP alleles based within 88 CFS-associated human genes using the SNP Genotyping GoldenGate Assay (Illumina, San Diego, CA, USA). Ancestry informative markers (AIM) were also examined.

    21 SNPs were significantly associated with CFS/ME, when compared with depression, & normal groups. 148 SNP alleles had a significant association with one or more CFS/ME subtypes. For each subtype, associated SNPs tended to be grouped together within particular genes. AIM SNPs indicated that 4 subjects were of asian origin while the remainder were Western European. Hierarchical clustering of AIM data confirmed the overall heterogeneity of all subjects.

    This study provides evidence that human SNPs located within CFS/ME associated genes are associated with particular gene expression subtypes of CFS/ME. Further work is required to develop this into a clinically useful aid to subtype-specific diagnosis.

    Comment by ME Research UK
    The information inherited from our parents (usually in the form of a gene, a sequence of DNA) has to be translated into a product, such as an RNA molecule or a protein, before it can be used by the body, and this process is called gene expression. Over the past few years, a number of research groups worldwide have investigated gene expression in people with ME/CFS, and the number of scientific reports published has steadily increased. Overall, the genes identified in the illness seem related to ‘‘immunity and defense’’, supporting what is already known about the role of the immune system.

    Dr Jonathan Kerr’s group at St George’s Hospital, University of London was one of the most active in defining the molecular basis of ME/CFS. Initially, the researchers performed a pilot study of gene expression in patients compared with controls, and have demonstrated marked human gene dysregulation, principally affecting the immune system. Their subsequent work on protein biomarkers (the backbone of any diagnostic test) identified several molecules which seemed to be specific to ME/CFS. In late 2007, the group published a paper (Journal of Clinical Pathology) outlining their identification of a putative “gene signature” for the illness consisting of 88 human genes. As the box below shows, these genes can be subdivided into categories by, say, diseases and disorders or by molecular and cellular functions. The research team said that three of the genes identified are directly linked with mitochondrial metabolism, and a further ten had indirect links with mitochondrial metabolism.

    Important disorders and functions associated with some of the genes in the putative ME/CFS gene “signature”


    Diseases: Haematological (22 genes), Immunological (14), Cancer (31), Dermatological (3), Endocrine system (9)

    Molecular & cellular function: Cellular development (26), Cell death (33), Gene expression (31), Cellular growth & proliferation (31), Cellular assembly & organisation (15)

    Physiological system development & function: Haematological system (22), Nervous, immune & lymphatic system (18), Tissue morphology (18), Survival (17), Immunity (20)

    As these 88 genes had been linked directly to the pathogenesis of ME/CFS, the logical next step was to begin the study of inherited determinants of susceptibility by examining single nucleotide polymorphisms (SNPs) – pronounced “snips” – within these genes. SNPs within some of these genes have been linked with features and complications which might be associated with CFS (eg. IL10RA SNPs are associated with development of lymphoma, a disease which some have speculated occurs increased frequency in CFS patients).

    With funding from ME Research UK, the St George’s group began work on identifying the key SNPs for each of the 88 genes. As there are hundreds of SNPs within each gene and the cost of studying all of them would be prohibitive, the team focussed on “determinative” SNPs (ie. those which are known to predict all or most of the others within one gene); the number of determinative SNPs per gene typically varies between 3 and 7 (mean = ~5). Once these determinative SNPs had been identified for each gene, the researchers designed a low density array cards to contain their respective assays, had these manufactured, and then used these to test genomic DNA samples of 105 patients in the initial sample group. After comparing allele frequencies between the CFS and normal groups, the allele frequencies were then be related to the gene expression levels for each gene.

    The results indicated that 21 SNP alleles could be significantly associated with CFS/ME patients as compared with Depression and Normal controls, and 148 SNP alleles which are associated with one or more CFS/ME subtypes. Assuming these results could be replicated, the authors believed that it might be possible to develop a subtype-specific diagnostic test using a subset of subtype-specific SNPs, to aid in the investigation and ultimately, the clinical management of CFS/ME patients.

    In 2011, a study was designed to assess whether the “gene signature” for ME/CFS (reported above) was sensitive enough to correctly classify a separate blind dataset of ‘mRNA relative quantities’ from a new population of ME/CFS patients and healthy persons. The results (published in PloS One) showed that the metric used was able to successfully classify roughly two-thirds of both ME/CFS and healthy samples, but – unfortunately – that the level of misclassification was high, leading the authors to conclude that the “gene signature” suggested in published scientific reports could not be used as a broad diagnostic test for ME/CFS.

    What are SNPs?
    Single nucleotide polymorphisms (SNPs) are small genetic changes in DNA that vary between individuals. In fact, humans are 99% identical as regards their gene sequences, and the 1% which remains is mostly accounted for SNPs, of which there are approx. 10 million in the human genome. This makes SNPs very useful – they can serve as helpful landmarks for population genetic maps, but their greatest importance is in biomedical research for comparing specific regions of the genome between groups or individuals with and without a disease.

    While most SNPs are silent, some can have important consequences for individual susceptibility to disease and reactions to treatment. An example is the apoE gene which is associated with an increased risk of Alzheimer’s disease. It is also thought that certain SNP combinations can contribute to a predisposition to developing medical conditions.

    At present, an enormous literature exists reporting possible associations between SNPs and diseases. for instance, the SNP500Cancer project is examining samples to locate important SNPs, and there are many such examples in a range of illnesses.

    The challenge in ME/CFS is to identify SNPs which will ultimately allow patients to be quickly and simply “diagnosed” from a sample, and possibly assigned to illness subgroups for specific therapies.
     
  3. Bob

    Bob

    Messages:
    9,846
    Likes:
    33,923
    England (south coast)
    Last edited: Sep 21, 2014
    cigana, snowathlete, Sea and 8 others like this.
  4. worldbackwards

    worldbackwards A unique snowflake

    Messages:
    2,091
    Likes:
    10,354
    Earth
    Bloody hell, when they send ME scientists into exile, they don't do it by halves, do they.
     
    taniaaust1, cigana, Sean and 5 others like this.
  5. Sidereal

    Sidereal Senior Member

    Messages:
    3,081
    Likes:
    17,069
    Oh my god.
     
    cigana, Nielk and Bob like this.
  6. Min

    Min Guest

    Messages:
    1,386
    Likes:
    1,924
    UK
    How many times was this fine doctor's gene expression research turned down by the UK's Medical Research Council, who then manage to waste £6.5 million of taxpayets' money on PACE and FINE? Their blatant bias in favour of funding psychobabble is scandalous.
     
  7. NK17

    NK17 Senior Member

    Messages:
    592
    Likes:
    1,516
    Dr. Montoya is from Colombia, makes me wonder if this is just pure coincidence? It probably is.
    Still we should make sure that they connect.
    At this point in time we need as many good researchers and clinicians' collaborations as we can get.
     
    merylg, MeSci, Kati and 2 others like this.
  8. Simon

    Simon

    Messages:
    1,913
    Likes:
    14,348
    Monmouth, UK
    I am not too impressed by this, because, as usual, there are concerns over replication:
    The problem is that this work never replicated, as Jonathan Kerr himself showed:
    Given the gene expression work didn't hold up, I'm not too sure why the new SNP work is expected to hold up better (albeit they targeted 'CFS-associated' genes). Nb out of 504 SNPs only 22 were significantly different in CFS vs controls.. What's really needed is to show they can detect the same differences and same subtypes in an independent sample - until then, we could just be looking at quirks in the data - which is what happened in the gene expression studies.
     
  9. alex3619

    alex3619 Senior Member

    Messages:
    12,285
    Likes:
    33,808
    Logan, Queensland, Australia
    The problem is false positives. In a large data set from a small number of samples you expect to see a lot of false positives. Further work is needed to see if any of these are true positives. It also depends on p values ... I hope they were tiny. A large p value cutoff for significance is behind so much medical babble. I have yet to read this paper.
     
    taniaaust1, Valentijn and Sean like this.
  10. user9876

    user9876 Senior Member

    Messages:
    2,436
    Likes:
    16,573
    I think its interesting that they are finding clusters. To my mind there are two different issues:
    1) As Simon says can these clusters be found in independent samples. And here it may be interesting to look if they are meaningful in terms of any subtype explanations.
    2) Are the clusters stable over time. That is taking multiple readings at different times for different patients - do they cluster in the same way. I seem to remember the Lights work looking at gene expression after moderate exercise. Hence there may be an issue around when samples are taken and hence a lack of stability in the results.
    The first point is easy to test but the second is much harder to understand since it means being very careful about when and what activities were done prior to the samples being collected.

    The last point worth making is that hierarchical cluster analysis is a nice technique but can be sensitive to the distance metric used particularly if clusters are not well separated. That's true in general for cluster analysis as well as issues with odd shaped clusters.
     
    snowathlete and Valentijn like this.
  11. alex3619

    alex3619 Senior Member

    Messages:
    12,285
    Likes:
    33,808
    Logan, Queensland, Australia
    I would be very surprised if all of our biochemical and gene expression patterns were stable over time. I guess that some are, but not all. However snps are not subject to such variance. When looking at snps, you are looking for risk factors. They might not be directly causal, or fully causal, but may contribute in some way.

    What is more likely to be unstable is the symptom based subtypes. I have possibly fit a bunch of different ones at different times over the decades.
     
    Last edited: Sep 22, 2014
    taniaaust1, Dolphin, merylg and 2 others like this.
  12. alex3619

    alex3619 Senior Member

    Messages:
    12,285
    Likes:
    33,808
    Logan, Queensland, Australia
    In case I was not clear, we know for a fact that our short term gene expression is variable in an atypical way, including post-exertional and circadian factors. I took that for a given in my last message.
     
    merylg, NK17 and MeSci like this.
  13. NK17

    NK17 Senior Member

    Messages:
    592
    Likes:
    1,516
    You couldn't say it better @alex3619!
     
  14. Dolphin

    Dolphin Senior Member

    Messages:
    10,574
    Likes:
    27,659
    I've just read the paper. My knowledge of genetics is not good and so some stuff I don't understand well.

    Looking at the SNPs listed in table 2 ["Single-nucleotide polymorphisms (SNPs) showing a frequency distribution that was significantly different between CFS/ME, depression and normal, and between CFS/ME and normal"], a lot of the 21 differences are due to the depression group being different to the others.

    11 of the 21 are listed as not significant in this column: "χ2 test, p value for distribution between CFS/ME and normals"

    I am also suspicious a few of the others are not significant either (i.e. there might be an error). For example for one gene allele, 5.6% of the CFS/ME group had it compared to 5.1% in the normals; for another, it was 4.9% versus 5.4% respectively.
     
    Last edited: Oct 20, 2014
    Valentijn likes this.
  15. alex3619

    alex3619 Senior Member

    Messages:
    12,285
    Likes:
    33,808
    Logan, Queensland, Australia
    There are two issues with "significant" results. p-value is a mathematical measure of the likelihood of relevance to a particular measure. Even experimental bias can generate a significant result.

    Someone with better knowledge of statistics and probability than I have might like to chime in.

    However its also indicative of the risk of false positives.If a large data set is examined then at about 5% p value you would expect about 5% false positives - this is not carved in stone, is not precise, but its indicative.

    In physics they want really low p values ... like 0.0001. 5% is really high, with too big of a chance of giving false results.

    Now let us look at the flip side of the issue. The data set probably represents a heterogeneous patient group. Its not one disorder. So you would not expect highly significant results. What you can only do is find candidates for further, and hopefully better and more focused, research.

    This is exploratory science. Its about asking questions and looking for ways to ask even better questions. Its not about definitive answers.
     
    Valentijn, Kati and NK17 like this.
  16. Snow Leopard

    Snow Leopard Hibernating

    Messages:
    4,578
    Likes:
    12,179
    South Australia
    I don't understand table 4, it states that the 'CFS/ME associated allele' are not the same as the 'Mutation resulting in no binding (predicted)', but yet they said:

    Is there an error, or am I reading it wrong?
     
  17. Bob

    Bob

    Messages:
    9,846
    Likes:
    33,923
    England (south coast)
    NK17 likes this.
  18. Little Bluestem

    Little Bluestem Senescent on the Illinois Prairie ❀❤✿Ƹ̵̡Ӝ̵̨̄Ʒ✿❤❀

    Messages:
    4,607
    Likes:
    5,193
    Midwest, USA
    When research is done with living organisms, the p-values will be much higher than is physics, because of the individual variability of living beings. It has been too long since I was involved with such things for me to remember what a reasonable p-value would be. It may also depend on what species you are doing research with.
     
    NK17, Sidereal and adreno like this.
  19. alex3619

    alex3619 Senior Member

    Messages:
    12,285
    Likes:
    33,808
    Logan, Queensland, Australia
    Yes, but its also a cause of huge numbers of false positives.
     
  20. nandixon

    nandixon Senior Member

    Messages:
    785
    Likes:
    1,892
    @Dolphin, Would you, or anyone else, please post a copy of Table 2 here? Thank you!
     

See more popular forum discussions.

Share This Page