Mission Software Resources CompMed EASDcourse Contact

Course topics

Monday
Large-scale studies of diabetes
Tuesday
Clinical characterization of diabetic complications
Wednesday
Dissection of genetics of diabetes and its complications
Thursday
From genes to function
Friday
Metabolic patterns of diabetes

Route to Biomedicum

Exercise 4: Nuclear proliferation

It is highly recommended that you go through the first two exercises before starting this one.

Nuclear magnetic resonance (NMR) spectroscopy is a non-invasive technique to measure a number of metabolites in a biofluid such as blood serum. In this exercise, a sample dataset of 613 individuals from the FinnDiane Study with type 1 diabetes is provided for analysis [Mäkinen et al. 2008].

Task 1: View the material

The dataset contains a number of NMR signals at different frequencies and clinical variables such as diabetes duration and kidney disease status. Data and detailed descriptions are available via the links below. Look carefully both the data and the configurations files: you will notice that all the spectral and biochemical variables are designated as training data (InputVariable) and all the clinical measures are independent tests (TestVariable). Put differently, we wish to know if the biochemistry contains intrinsic statistical structure that can be associated with the clinical features.

Download data
Download config/info

Task 2: Correlation structure

Use the upload form to submit the data file to the Katiska tool.

Go to Katiska

Leave the result page open in the browser. You should also download and save the result archive as 'nmr_correlation' onto your desktop for later viewing.

Task 3: Metabolic profiles

For the Melikerion tool, you need also the configuration file. Submit the necessary files to the online system.

Go to Melikerion

Task 4: Review results

You should now have both the correlation network and the self-organizing map of the NMR dataset available. Compare the two analyses and determine the most significant NMR variables and their associations to the clinical features.

Questions

  1. In the correlation network, are the clinical features and metabolites strongly inter-connected? Are there connections that can be explained by the clinical definition rather than a biological interaction?
  2. What are the most important NMR signals with respect to the SOM structure?
  3. What is the most significant clinical marker?
  4. Are there any biological basis for the findings?
  5. Type 1 diabetic patients with kidney disease have often high triglycerides but low HDL cholesterol. Lipids produce a complex NMR signal when they reside inside lipoprotein particles, but can you verify this finding from the available NMR signals and map colorings?
  6. Is glucose a strong determinant of the map structure? Is this a good or a bad thing?
  7. Compared to the previous exercises, are the results easier or more difficult to interpret? What aspects of the results influence interpretability?

GWAS exercises

Download material

Statistics exercises

1) Networking without Facebook

2) Some are more equal than others

3) Textbook case

4) Nuclear proliferation

Updated 2009-11-27 by vpmakine.