The steps of text mining in biomedical field and the methods used in its each step were described with stress laid on the tools used in each step of text mining in order to promote text mining in biomedical field.

The long-year program for medical education is the major way to train the high level medical professionals.This paper colleeted the research papers on this topic published in Chinese journals in recent years,extracted their key words and counted the high frequent key words According to the co-occurrence of a pair of key words in one paper.these high frequent keywords were clustered into groups.The research structure and main directions of the long-year program were outlined as the study of macro-planning of long-year program,the study of teaching models and teaching system,study of teaching practice.These results could be a useful reference for the medical educational researchers and managers.

After the basic properties of literature-related citation network, co-authorship network and co-words network were analyzed and the advances in their application research were summarized in aspects of their construc-tion methods, size and research depth, it was pointed out that article similarity networks could be constructed using the article similarity algorithm, and their basic properties and features were analyzed.

The paper introduces using MedlineR to carry out data mining from Medline, for observing the gene relation, and using Pajek to realize the visualization of the results. Practice proofs that MedlineR could help researchers come up with assumptions in the field that they are interested in, and get a visual fingerprint. These results are the bases to go to more complex literature data mining and an important supplementary to experimental methods.

Co-citations of highly cited papers on gene annotation covered in Web of Science were analyzed by clustering analysis using clustering software after the word matrix of resource literature and highly cited papers was formed, which showed that the application of text mining on gene annotation includes use of authorized tools, development of text mining tools and algorithms, and verification of text mining tools.

Co-citations of foreign and domestic highly cited papers on drug target discovery were analyzed by clustering analysis using BICOMB2.01 and gCLUTO.Semantic analysis of the titles and abstracts in these highly cited papers and their important source literature showed that general trend, theoretical foundation, main methods and principal resources are the major hotspots of text mining in drug target discovery.

Text mining provides a new approach for drug repositioning, and the emerging databases and their corresponding tools provide more and more convenience for drug repositioning by text mining.The methods and tools for studying drug repositioning by text mining and their successful application examples were thus described in this paper in order to provide reference for the researchers interested in this field.

OBJECTIVE To investigate the bacterial distribution and the extent of drug resistance in lower respiratory tract disease among aged people.METHODS From Jan 2006 to Dec 2006 we collected 304 sputum samples of the old patients who were diagnosed lower respiratory tract infection,bacterial distribution and antibiotic resistance of 178 strains of the pathogens were analyzed retrospectively.RESULTS The most common four kinds of Gram-negative bacilli were Pseudomonas aeruginosa,Klebsiella pneumoniae,Escherichia coli and Acinetobacter baumannii.For above bacteria,the ratio of the resistance to ampicillin,cefazolin and compound sulfamethoxazole was over 40%.The isolating rate of ESBLs-producing E.coli and K.pneumoniae was 29.8%.CONCLUSIONS Monitoring the common bacteria and ESBLs bacteria precisely can provide therapy,guide to early cure and control lower respiratory tract infection of aged patients effectively.

AIM: To determine the aqueous, vitreous, serum levels of pigment epithelium-derived factor (PEDF)and vascular endothelial growth factor(VEGF)in patients with proliferative diabetic retinopathy(PDR), and to speculate on the source of the change in concentration and to discuss its clinical significance.METHODS:Forty-one eyes withproliferative diabetic retinopathy were included in the study, 16 of which were complicated by neovascularization of iris (NVI).Twenty-one eyes with idiopathic macular hole (MH)were as controls. The aqueous, vitreous, serum levels of PEDF and VEGF of all the groups were determined with ELISA. PEDF, VEGF and the levels in the three groups were compared with analysis of variance (ANOVA). The PEDF, VEGF concentrations in aqueous,vitreous and serum were analyzed with Pearson correlation test,and the correlation of PEDF and VEGF levels was also analyzed with Pearson correlation test.RESULTS:The aqueous levels of PEDF decreased significantly in sequence in groups of control, PDR without NVI, PDR with NVI. VEGF levels increased coordinately. The similar findings existed in vitreous samples. The PEDF,VEGF levels in aqueous were not correlated significantly with those in serum,but correlated positively with those in vitreous. The intraocular levels of PEDF had a negative correlation to those of VEGF.CONCLUSION:The reduction of intraocular PEDF level and elevation of intraocular VEGF level may play an important role in the occurrence and progression of PDR. In the development of PDR, the PEDF,VEGF levels in aqueous may be mainly effected by local pathological changes, as anti-angiogenic and pro-angiogenic factors, their unbalanced intraocular distribution may promote the angiogenesis of the iris and retina.

A paper similarity network was constructed in light of semantic similarity algorithm using the complex network processing package , igraph in R language , and analyzed by random walk-trap algorithm , label propagation algorithm, BGII algorithm, and Girvan-Newman algorithm, respectively.The accuracy and stability of these 4 al-gorithms were compared according to the golden standards and the D function for network community classification evaluation index, which showed that the accuracy and stability of random walk-trap algorithm were better than those of the other 3 algorithms and preconditioning of complex network was an important influencing factor for clustering .