Gene Disease Cooccurrence In Literature

Question

[Zhe Liu]

Dear all,

I want to investigate the correctness of gene discovered in microarray dataset. So I want to have a look at the cooccurrence of these gene and specific disease, say cancer, in Pubmed literatures. Do you know some relevant R packages or web services could meet this end?

Thanks a lot in advance and I am looking forward your reply.

Zhe Liu

[Larry_Parnell]

Zhe Liu: Thank you for replying to the responses below. Just a reminder to vote up those responses that you find helpful and to assign one as the accepted answer.

Answer 1

[Madhan]

Apart from the text mining tools to extract information from the Pubmed literatures, there are few High quality Databases that can be used for this purpose.

I mostly use GeneSigDB (Harvard) and Genetic Association Database (GAD from NCBI)

1. http://compbio.dfci.harvard.edu/genesigdb/

[Zhe Liu]

Thanks Madhan, this site is easy to use and I will try to upload some results and see what happens:)

[Madhan]

You are welcome :)

Answer 2

[Ivanka Karageorgieva]

Hello Zhe Liu,

I could offer you three alternative decisions to your problem:

1. PhenomicDB - it's a free multi-organism phenotype-genotype database unifying a variety of primary sources to make phenotypic data from a wide range of species and model organisms simultaneously searchable, visible and comparable. You can type your gene of interest (or even use a particular ID number of a particular DB if you are aware of any using the "External IDs" index in the sections drop-down) and all reported phenotypes, which are usually some king of diseases, that have relationship with your gene will be shown in the results. OR you can search by a particular disease/phenotype and see what genes are related to it. Say if you need any support.
2. Drug2Gene - it's a free database and unifies a number of popular public resources to provide structured and organized information for identified and reported relationships between genes/proteins and drugs/compounds. Here the disease information is indirectly related to the result entries - it is contained in the Comments of the Activities column of the results table. You can search for a particular gene (by Official Gene Symbol, or Gene Name, or a particular ID number if you are aware of some, or any index of interest) and in the second search field to type the disease of interest choosing the Activity Comment index, then search for relations. Note that the results will actually show a relation between a gene and a drug, but the disease info will be available in the Comments of the Activities column. The second way is to search for a particular disease choosing the All Text index or the Activity Comment index, again – the results will show genes (of course related to drugs, which you can ignore), but also related to the particular disease that you typed in the query - the disease info will be available in the Comments of the Activities column, again.
3. Text mining – this method has already been suggested in the above posts. It is the most direct way to find co-occurrences of two terms or biological types (e.g. gene and disease) in the PubMed literature. You can use this tool: http://webrt.metalife.com/ in case you have particular abstracts of texts to paste – you can search for your terms of interest in the two main categories: Genes/Peptides/Proteins (may be Enzymes, too) and Diseases – found terms from each category would be highlighted in the text and you can investigate the co-occurrences in the text. However, if you don’t have PubMed literature texts available, but you’d rather prefer to search for some that deal with your terms/categories of interest, then an NCBI online tool for text mining will be more appropriate, because they are designed to make the analysis within the entire set of PubMed publications. Try some of these:
   • http://www.nactem.ac.uk
   • http://gate.ac.uk
   • http://www.ontotext.com/kim/
   • http://www.h-techsight.org
   • http://www.kmi.open.ac.uk/projects/magpie

Answer 3

[Sean Davis]

You could use eUtils from NCBI to query against pubmed (with a sleep in the loop so as to not get banned by NCBI) over all genes one-at-a-time with your disease term included. There was a tool called medliner that is no longer supported (I don't think) that did this kind of thing automatically.

If you are working in cancer, then I suggest you review several cancer-related databases. This site includes a short list that is not at all inclusive but might get you started. Finally, take a look at this project:

https://wiki.nci.nih.gov/display/ICR/Cancer+Gene+Index+End+User+Documentation

[Zhe Liu]

Thanks a lot Sean, I think the link https://wiki.nci.nih.gov/display/ICR/Cancer+Gene+Index+End+User+Documentation is a good start to learn cancer genomics databases.