I have two sets of GSE from GEO database and I would like to develop a prediction model in one set and velidate the model in another set. How can I match the gene names? The code are enclosed below.
GSE65682=getGEO('GSE65682',destdir = '/Users/zhang/Documents/2020/GEOsepsis/Data', AnnotGPL = T,GSEMatrix = T) if (length(GSE65682) > 1) idx
It is just a classification problem, that may need non-parmaetric regression, SVM etc methods. See e.g., https://academic.oup.com/bioinformatics/article/22/1/88/218231
Conference Paper Using deep learning to enhance cancer diagnosis and classification
https://www.ncbi.nlm.nih.gov/pubmed/30611843
As mentioned above, you can use the annotation packages from Bioconductor. In the GSE files you have the expression values relative to the probes in the microarray, these can be very different from one microarray to another one. You have to check from GEO database which microarray was used and apply the corresponding annotation package to get a human-readable name.
You can choose as suggested SYMBOL, which is the gene name, or others (ENSEMBL_ID, for transcript names).
Keep in mind that if the microarrays used are different, also the investigated transcripts/genes can be different (older microarrays usually have less resolution). In this case, you cannot expect to retrieve the same list of features at the end of the annotation process. You could just remove the non-matching features if you loss few features. Another possibility is to use an algorithm to impute missing values in the data (e.g. K-Nearest Neighbours imputation).