How to get the list of genes that could predict diagnosis
Hi there, I just got a question. We got some PBMC samples from a rare disease (10 samples). I have done the micro-array and found a set of genes that are deferentially expressed compared with the healthy controls. So my question is can I predict a set of genes that could help diagnosis? Thanks!
• 2,444 views
•
link
updated
by
natasha.sernova
•
written
by
znh.bing.liu •
•
written
by
znh.bing.liu •
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
Expression of genes in KEGG pathway
written by nhaus •
-
Correlation among genes in micro array data
written by izsyed16 •
-
What should be noted if analyzing the combination of lab-producing RNA-Seq data and open RNA-Seq da…
written by Zhilong Jia
-
gene expression in disease and normal
written by rob.costa1234
-
number of grouping affect log2 fold change in DESeq2 analysis?
written by star
-
GEO study selection for analysis in R
written by 786
-
Differential lncRNA-mRNA co-expression
written by hana
-
Up and Down regulated genes
written by mathavanbioinfo
-
What should be the frequency of a Single Nucleotide variation to be called as pathogenic variant fo…
written by garimav89 •
-
Working with GTEX dataset
written by pbio
These authors compared gene expression for ill and healthy people,
but they have already suspected some diseases.
https://www.jci.org/articles/view/59255
Hi ,I will take a through look at this paper!
You have to clarify what you mean by "help diagnosis". You already have your genes that are altered in the disease.
Hi igor,Thank for your reply! As to "help diagnosis", I would like to find out some genes (eg, 20 or so), and draw a ROC curve dependent on these genes. The combination of genes could help improve the diagnostic sensitivity and specificity. Thanks!
Why not just try the top most significant genes?
If you want something more advanced, how about k-Nearest Neighbour? https://stat.ethz.ch/R-manual/R-devel/library/class/html/knn.html
The problem is that we have only got 10 patient samples. So I am wondering whether it is enough to perform a ROC curve analysis to determine the diagnostic sensitivity and specificity?
I will give you a citation from the paper I recommended.
"We prospectively recruited 35 patients with active CD and 32 patients with active UC, prior to commencing treatment.
Of these patients, 58% were newly diagnosed (23 out of 35 patients with CD and 16 out of 32 patients with UC)".
For their purposes it was enough. You can go to KEGG, input the rare disease name and you will definitely get more
involved genes.
See this post, it's very useful. There should be an updated version, I will look.
A: Get The List Of Genes Associated With A Disease
This is another useful post. David is not good sometimes, they lost something while updated it,
but other databases are OK as far as I know.
A: Prediction Causative Gene By Network
See also these posts:
gene (or SNP) related to a disease
GWAS for rare diseases?
This post allows to search biomedical literature:
A: Mark2Cure launches tool to help crowd-source curation of the biomedical literatu
Linking Genes/Diseases To Pubmed Ids
This site may also be useful:
http://massgenomics.org/2013/04/rare-variants-human-disease.html
It really depends on your particular field. Check other similar studies and see what they do.
At the very least, you can try leave-one-out cross-validation. Some additional info:
http://stats.stackexchange.com/questions/27454/how-does-leave-one-out-cross-validation-work-how-to-select-the-final-model-out
http://www.russpoldrack.org/2012/12/the-perils-of-leave-one-out.html