Jianqiang Du

Mass spectrometry-based proteomic analysis of Kashin-Beck disease

Kashin-Beck disease (KBD) is a degenerative osteoarticular disease of unknown etiology. The management of KBD would benefit from the identification of the biomarkers related to this disease. In this study, mass spectrometry (MS)-based proteomic profiling was used to identify potential biomarkers of the disease. One hundred and sixteen serum samples of KBD cases and healthy controls were collected and analyzed. A framework for data analysis was implemented, which included normalization, denoising using undecimated discrete wavelet transforms, baseline subtraction, peak detection and alignment, non-parametric testing and classification by support vector machine. The method identified correlative mass points and obtained a discriminative pattern with 90.91% sensitivity and 82.61% specificity. The results of this study, although preliminary, suggest that further proteomics study may be useful with a larger number of appropriate specimens, careful experiment manipulation and improved MS techniques.

A Biological Term Based Information Retrieval Web Service that Assists the Understanding of Biology Literatures

At present, there are many professional terms and database record identifiers in biomedical literature. When a lot of unknown terms appear in the literature, it brings trouble in understanding the document. In this paper, a biological terminology information retrieval web service was established. First, a terminology database was constructed, whose records were come from databases supported by BioWarehouse, as well as other biological databases. Then, the server side code of the web service was developed to realize the searching and communication function, which transmit messages between client and server through HTTP and SOAP protocol. Based on the professional terminology database, this web service can be used to query unknown biological terms appearing in literature and accurately show its explanations. It facilitates researchers to understand scientific literatures in biological field.

Biomarker Discovery from Proteomic Data Based on Wavelet Package Transform and AdaBoost

Advances in proteomics provide a new method for early detection of cancer, it can provide a wealth of information and rapidly generate large quantities of data from the analysis of biological specimens. In particular, proteomic pattern of body fluid has attracted attention as an approach to early detection of cancer. Mass spectrometry can provide rapid and precise measurements of the proteins in the body fluid. But the data processing is still a challenge due to noise artifact and high dimensionality of the proteomic data. In this paper, we proposed a scheme that combined wavelet package transform, statistic analysis and AdaBoost to process a public prostate cancer proteomic dataset, the obtained discriminative proteomic pattern can differentiate the cancer form control with high sensitivity and specificity.

Wavelet Transform and Bagging Predictor Approaches to Cancer Identification from Mass Spectrometry-Based Proteomic Data

The early detection of cancer has the potential to dramatically reduce the mortality of cancer. Recently, using the mass spectrometry based proteomics to develop profiles of patient serum proteins, combined with bioinformatics algorithms has been reported as a promising method to achieve this goal. In this paper, we develop a workflow that combined wavelet transform, statistic analysis and bagging predictor to process a public ovarian cancer proteomic dataset, and ultimately obtained a discriminative proteomic pattern that can differentiate the cancer form control with high sensitivity and specificity. Compared with the previous studies, the results of our study are based on peaks of mass spectrometry and the discovered discriminative pattern is more biologically.

Common Features Identification of Differently Expressed Genes Related to Endemic Osteoarthritis Disease

Kashin-Beck disease (KBD) is a chronic, endemic osteochondropathy in China. cDNA microarray techniques were used to select the differentially expressed genes between the healthy and KBD patients. 227 differently expressed genes were used to identify common characteristics. By using GeneMerge anaysis, the common functions of these genes were identified and some of which are related to metal ion binding, such as magnesium ion binding and zinc ion binding. The upstream sequences of these genes were also downloaded from genomic databases and a sequence dataset was constructed. A distance tree was then achieved according to sequences similarity comparing. Finally, 7 gene clusters, each of which contained 2~3 genes with similar function were obtained.

Classification of CpG Islands in the Human Genome Based on the Interval Distance Distribution of Adjacent CG Sites

There have been many studies analyzing relations between CpG islands and gene functions. Most results showed that promoters of many housekeeping genes contain CpG islands, however, the relation between gene functions and CG dinucleotides positions in CpG islands was less considered. In this study, we try to classify CpG islands according to interval distance distribution of adjacent CG sites and find some functional correlations. First the human genome sequences were downloaded from the EMBL Nucleotide Sequence Database. Then a dataset was constructed, each record of which is an interval distance distribution of adjacent CG sites of a CpG island. Finally an algorithm was designed, which can calculate approximately minimal difference of any two records. Based on the algorithm, we obtained many classes using the hierarchical clustering method, each of which contains some similar CpG islands, and some of their common features were studied.

Preprocessing and Classifying of Mass Spectrometry-Based Proteomics Data Using Wavelet Transform and Decision Tree Learning

Advances in mass spectrometry-based proteomics have brought expectations for biomedical researchers. It can be used for identify proteomic patterns in body fluids to discriminate patients from control, the results are inspiring. However, most of the earlier studies are based on the direct application of original MS data, together with dimension reduction or feature selection methods. We deemed that only the peaks of MS data have real biological meaning, so its important to obtain the ultimate proteomic pattern using the real peaks. In this paper, we proposed a workflow that combined wavelet transform, statistical analysis and decision tree learning to process MS data. Especially, the statistical analysis which have not been attached too much importance in most studies was investigated, the possible distribution law of the MS peaks was proposed.