Changes between Version 6 and Version 7 of Courses/ComputationalMolecularBiologyResearch2015/P9

Timestamp:

2015-02-04T14:45:22+01:00 (10 years ago)

Author:

mdijkstra

Comment:

--

Courses/ComputationalMolecularBiologyResearch2015/P9

@@ -11 +11 @@
 Patrick Deelen
 == Introduction ==
-Different molecules in your body can react with glucose. Some of the resulting molecules are called AGEs (Advanced Glycemic End products). These AGEs accumulate during life span in different organs, including skin. So, the amount of AGEs is related to one's age (i.e., years since birth). Other factors, however, may accelerate the accumulation of AGEs. These factors include smoking, and also age related diseases like diabetes. The AGE molecules appear to be auto-fluorescent. Based on this property, UMCG has developed a novel, non-invasive screening method that can quantify the amount of AGEs in one's skin. Two huge data sets are created with a lot of phenotypic patient information, including their autofluorescence scores!
+Different molecules in your body can react with glucose. Some of the resulting molecules are called AGEs (Advanced Glycemic End products). These AGEs accumulate during life span in different organs, including skin. So, the amount of AGEs is related to one's age (i.e., years since birth). Other factors, however, may accelerate the accumulation of AGEs. These factors include smoking, and also age related diseases like diabetes. The AGE molecules appear to be auto-fluorescent. Based on this property, UMCG has developed a novel, non-invasive screening method that can quantify the amount of AGEs in one's skin and thereby quantify your 'internal age'. Two huge data sets are created with a lot of phenotypic patient information, including their autofluorescence scores!
 == Project 9 - Optimising decision trees to differentiate between healthy and diseased using parameter optimisation ==
 Different scientific articles have been produced on this topic. One of those articles presents a 'decision tree' that can, and is, used by doctors to predict risk on (pre-)diabetes. Reproducing this decision tree is the first step in this project. As a next step you will come up with a criterion that indicates the performance of a given decision tree. Next, you will optimise that criterion as function of the cut-offs in the tree. We challenge you to come up with an even better decision tree. Do you realise that information may be lost each time the tree branches? Maybe you can come up with a model that does not have this drawback? You will also answer questions about the reproducibility of your research. To what extend are your finding generalisable from one data set to another one?