5 ms·

I looked at the paper and the data is kind of strange. Looking at figure 2, there are a bunch of cytokines that have a nice linear progression with increasing l

by kens 9y ago

I looked at the paper and the data is kind of strange. Looking at figure 2, there are a bunch of cytokines that have a nice linear progression with increasing levels corresponding to increasing severity of disease. This seems like solid evidence that more cytokine = bad. But if you compare with the controls, mild cases of ME/CFS have less than the control, moderate cases have the same as controls, and severe cases have more. This kind of torpedoes any simple explanation of what's going on. The authors discuss this in a kind of hand-wavy way that maybe ME/CFS causes some down-regulation to start with and then the increasing trend is on top of that. So, I'm not sure what to make of this. It seems useless as a test at this point: "You have totally normal cytokine levels, indicating that you're either healthy or have moderate ME/CFS."

Gibbon1 9y ago
The thing that currently spazs me out is the Simpson Paradox (Yule–Simpson effect) https://en.wikipedia.org/wiki/Simpson%27s_paradox https://en.wikipedia.org/wiki/Simpson%27s_paradox In short, if your population has two (or more) diseases with similar enough symptoms that they are grouped together, statistical analysis can and does totally fail. And your experiment will produce garbage. So what if Chronic Fatigue Syndrome is really caused by a couple of different things, but with converging symptomology. You have a real catch 22 situation. Can't identify the cause of the disease without a consistent experimental group. Can't select a consistent experimental group without a functional test.
- kens 9y ago
  That's a very good point. The paper suggests that "severity may be a key variable for subgrouping ME/CFS." So it's possible that there are three different ME/CFS diseases that can be distinguished by cytokine levels. Hopefully once these are separated, statistical analysis will succeed and medical progress would follow. In the bigger picture, you have to wonder how many things that we currently categorize as one disease are really many different diseases. (Like how in the 1800s, people would die of "a fever" but we now know there are many different diseases causing fevers.) It's kind of a cliche now that cancer is not a single disease. I expect conditions such as autism and schizophrenia will fragment once the right tests are available. What else?
  - Gibbon1 9y ago
    > how many things that we currently categorize as one disease are really many different diseases. Had a GI doctor tell me he thinks there are 'lots' of gastrointestinal syndromes that medicine has no way to identify. Also have read an epidemiologist years ago that wrote there are probably a lot more infectious diseases out there. Some of which may have serious long term complications.
    - tomkinstinch 9y ago
      Fortunately we have a tool to help on both fronts: metegenomic sequencing. It's only just starting to be used in clinical settings in a non-retrospective way, but many groups are interested in how the gut microbiome (or changes of) can impact the etiology of GI disease. Several pharma startups are looking at therapies for GI disease that rely on changing the gut microbiome, either directly via introduction of new flora (fecal transplant or introduction of tailored microbial cocktails), or via small molecules that can shape the taxonomic distribution. Microbial cultures introduced to the GI tract tend to either die off entirely, or become dominant. It's a bit of an unsolved puzzle why the trend isn't toward plurality. The microbial therapies for C. diff. are one example. Beyond GI disease there's also indication that the gut microbiome has psychological impact, and metabolic impact. There's a lot we don't know. Metagenomic sequencing can disentangle the various organisms present in a sample and report all of the taxa present, but the results are only as good/complete as the database used, and unknowns can be a significant slice of the pie. When our group does deep-enough sequencing and de novo viral genome assembly, we often have enough contigs beyond our organism of interest to assemble genomes for other viruses, and sometimes they are novel. Yes, there are more infectious diseases out there. A Chinese team exploring the viruses of invertebrates found nearly 1,500 new viruses [1]. That one study expanded the list of known viruses by a significant number, we're talking double-digit percentage change. Many more viruses will be found, and some of them will be pathogenic. In the (hopefully) near future it will be reasonable (if not yet cost effective) to use tools like high-throughput sequencing to look at everything in a sample, rather than ordering specific assays for individual pathogens. It's a long road until sequencing will be part of the standard of care, but we're in a golden age of biomedical exploration, enabled by sequencing and new molecular methods. 1. http://www.nature.com/nature/journal/v540/n7634/full/nature20167.html http://www.nature.com/nature/journal/v540/n7634/full/nature2...
- Terr_ 9y ago
  Makes me think of Anscombe's Quartet. https://en.wikipedia.org/wiki/Anscombe%27s_quartet https://en.wikipedia.org/wiki/Anscombe%27s_quartet
  - Celine_ 9y ago
    I quit working at shoprite and now I make $35h - $80h...how? I'm working online! My work didn't exactly make me happy so I decided to take a chance on something new…after 4 years it was so hard to quit my day job but now I couldn't be happier. ........... http://ow.ly/iBXm30dNtIZ http://ow.ly/iBXm30dNtIZ
  - Gibbon1 9y ago
    I love Anscombe's quartet.
- CuriouslyC 9y ago
  It isn't really that much of a problem. If you collect data on a large enough cohort of symptomatic individuals, you can recover separate subtypes through cluster analysis.
  - dotancohen 9y ago
    Cluster analysis will require thousands of not tens of thousands of cases to study. And since we don't know even how many diseases there are, this will have to be an unsupervised CA, so for each case you'll be looking tens if not hundreds of parameters. Machine Learning techniques are well suited to crunching this type of data, but don't underestimate the difficulty in collecting so much data.
    - CuriouslyC 9y ago
      You don't need thousands of cases, but you probably do need on the order of around a thousand samples. A hundred individuals with ten samples each over time isn't that challenging. I've done it successfully with mouse work at that scale. I've found that in biological systems, individual variation can be as large as population variation, particularly for something "reactive" like the immune system, which is tied into stress response.
      - mrow84 9y ago
        Don't you have to be concerned about differences between the structure of individual variability as opposed to population variability? Perhaps for the distinctions you are trying to draw it is not an issue?
        CuriouslyC 9y ago
        I haven't found that to be a big issue. I choose clustering and dimensionality reduction hyperparameters to penalize clusterings with few repeated individuals (high entropy), while trying to maximize the entropy of the distribution of cluster labels. To prevent cluster count explosion, you can put a regularization term/prior on the total number of clusters. An exponential or gamma with the expectation set to the square root of the number of distinct individuals is a good starting point. Of course a big part of any analysis is responding to the data, no single approach will work every time.
- zik 9y ago
  Chronic Fatigue Syndrome is a syndrome - ie. "a group of symptoms which consistently occur together". The fact that it's called a syndrome rather than a disease is a recognition that it's not understood and may be in fact a grouping of different diseases with similar symptoms.
- Shikadi 9y ago
  Do you know if this has been accounted for with ADHD anxiety and or depression? It has always seemed to me like they represent symptoms rather than the actual cause, especially now that we know serotonin deficiency doesn't cause depression

by kens 9y ago

Gibbon1 9y ago

The thing that currently spazs me out is the Simpson Paradox (Yule–Simpson effect) https://en.wikipedia.org/wiki/Simpson%27s_paradox https://en.wikipedia.org/wiki/Simpson%27s_paradox In short, if your population has two (or more) diseases with similar enough symptoms that they are grouped together, statistical analysis can and does totally fail. And your experiment will produce garbage. So what if Chronic Fatigue Syndrome is really caused by a couple of different things, but with converging symptomology. You have a real catch 22 situation. Can't identify the cause of the disease without a consistent experimental group. Can't select a consistent experimental group without a functional test.

kens 9y ago
That's a very good point. The paper suggests that "severity may be a key variable for subgrouping ME/CFS." So it's possible that there are three different ME/CFS diseases that can be distinguished by cytokine levels. Hopefully once these are separated, statistical analysis will succeed and medical progress would follow. In the bigger picture, you have to wonder how many things that we currently categorize as one disease are really many different diseases. (Like how in the 1800s, people would die of "a fever" but we now know there are many different diseases causing fevers.) It's kind of a cliche now that cancer is not a single disease. I expect conditions such as autism and schizophrenia will fragment once the right tests are available. What else?
- Gibbon1 9y ago
  > how many things that we currently categorize as one disease are really many different diseases. Had a GI doctor tell me he thinks there are 'lots' of gastrointestinal syndromes that medicine has no way to identify. Also have read an epidemiologist years ago that wrote there are probably a lot more infectious diseases out there. Some of which may have serious long term complications.
  - tomkinstinch 9y ago
    Fortunately we have a tool to help on both fronts: metegenomic sequencing. It's only just starting to be used in clinical settings in a non-retrospective way, but many groups are interested in how the gut microbiome (or changes of) can impact the etiology of GI disease. Several pharma startups are looking at therapies for GI disease that rely on changing the gut microbiome, either directly via introduction of new flora (fecal transplant or introduction of tailored microbial cocktails), or via small molecules that can shape the taxonomic distribution. Microbial cultures introduced to the GI tract tend to either die off entirely, or become dominant. It's a bit of an unsolved puzzle why the trend isn't toward plurality. The microbial therapies for C. diff. are one example. Beyond GI disease there's also indication that the gut microbiome has psychological impact, and metabolic impact. There's a lot we don't know. Metagenomic sequencing can disentangle the various organisms present in a sample and report all of the taxa present, but the results are only as good/complete as the database used, and unknowns can be a significant slice of the pie. When our group does deep-enough sequencing and de novo viral genome assembly, we often have enough contigs beyond our organism of interest to assemble genomes for other viruses, and sometimes they are novel. Yes, there are more infectious diseases out there. A Chinese team exploring the viruses of invertebrates found nearly 1,500 new viruses [1]. That one study expanded the list of known viruses by a significant number, we're talking double-digit percentage change. Many more viruses will be found, and some of them will be pathogenic. In the (hopefully) near future it will be reasonable (if not yet cost effective) to use tools like high-throughput sequencing to look at everything in a sample, rather than ordering specific assays for individual pathogens. It's a long road until sequencing will be part of the standard of care, but we're in a golden age of biomedical exploration, enabled by sequencing and new molecular methods. 1. http://www.nature.com/nature/journal/v540/n7634/full/nature20167.html http://www.nature.com/nature/journal/v540/n7634/full/nature2...
Terr_ 9y ago
Makes me think of Anscombe's Quartet. https://en.wikipedia.org/wiki/Anscombe%27s_quartet https://en.wikipedia.org/wiki/Anscombe%27s_quartet
- Celine_ 9y ago
  I quit working at shoprite and now I make $35h - $80h...how? I'm working online! My work didn't exactly make me happy so I decided to take a chance on something new…after 4 years it was so hard to quit my day job but now I couldn't be happier. ........... http://ow.ly/iBXm30dNtIZ http://ow.ly/iBXm30dNtIZ
- Gibbon1 9y ago
  I love Anscombe's quartet.
CuriouslyC 9y ago
It isn't really that much of a problem. If you collect data on a large enough cohort of symptomatic individuals, you can recover separate subtypes through cluster analysis.
- dotancohen 9y ago
  Cluster analysis will require thousands of not tens of thousands of cases to study. And since we don't know even how many diseases there are, this will have to be an unsupervised CA, so for each case you'll be looking tens if not hundreds of parameters. Machine Learning techniques are well suited to crunching this type of data, but don't underestimate the difficulty in collecting so much data.
  - CuriouslyC 9y ago
    You don't need thousands of cases, but you probably do need on the order of around a thousand samples. A hundred individuals with ten samples each over time isn't that challenging. I've done it successfully with mouse work at that scale. I've found that in biological systems, individual variation can be as large as population variation, particularly for something "reactive" like the immune system, which is tied into stress response.
    - mrow84 9y ago
      Don't you have to be concerned about differences between the structure of individual variability as opposed to population variability? Perhaps for the distinctions you are trying to draw it is not an issue?
      - CuriouslyC 9y ago
        I haven't found that to be a big issue. I choose clustering and dimensionality reduction hyperparameters to penalize clusterings with few repeated individuals (high entropy), while trying to maximize the entropy of the distribution of cluster labels. To prevent cluster count explosion, you can put a regularization term/prior on the total number of clusters. An exponential or gamma with the expectation set to the square root of the number of distinct individuals is a good starting point. Of course a big part of any analysis is responding to the data, no single approach will work every time.
zik 9y ago
Chronic Fatigue Syndrome is a syndrome - ie. "a group of symptoms which consistently occur together". The fact that it's called a syndrome rather than a disease is a recognition that it's not understood and may be in fact a grouping of different diseases with similar symptoms.
Shikadi 9y ago
Do you know if this has been accounted for with ADHD anxiety and or depression? It has always seemed to me like they represent symptoms rather than the actual cause, especially now that we know serotonin deficiency doesn't cause depression