How to Build CHD@ZJU
CHD related Articles were retrieved from Pubmed, by entering keywords "coronary heart disease" and constrict the publish date from 2000/1/1 to now (2013/1/23).
As a result, totally 115898 articles were found and their abstracts were downloaded for text mining.
Since some articles didn't contain abstracts, only 88396 abstracts remained.
The text-mining process to get CHD related genes could be divided in to 5 following steps:
1) Extracting all keywords from abstracts and ignoring those keywords start with numbers. 101402 keywords were extracted.
2) Input these keywords into Gene library in ArrayTrack and find possible related genes. 4674 genes were then found.
3) Put these 4674 genes again into pubmed abstracts to find related aticles. Only genes which offical name or there keyword description (such as prolactin for gene PRL)
could be found in the abstract would be remained. As a result, 1247 genes were remained.
4) Manually examined on the 1247 genes to validate it was acutally related to CHD. Some genes would be filtered if it represents other meanings
(such as gene CAD, Entrez ID:790, carbamoyl-phosphate synthetase 2, is mostly meant coronary arterial disease in articles).
660 genes were then validated with at least one reference.
5) All genes was compared with 1078 CHD genes in RGD database, and 370 genes were overlapped.
These 370 genes were labels as "RGD_Supported" and the other 290 genes were labels as "REFERED".
All 663 genes had supported references in CHD@ZJU which were examined by step 4.
How To contact Us
Collaboration Information: Prof. Xiaohui Fan (firstname.lastname@example.org)
Website using assistance : Leihong Wu (email@example.com)