Waiming Kong | PubMed Health

PubMed2007

Predicting single nucleotide polymorphisms (SNP) from DNA sequence by support vector machine.

Recently, SNP has gained substantial attention as genetic markers and is recognized as a key element in the development of personalized medicine. Computational prediction of SNP can be used as a guide for SNP discovery to reduce the cost and time needed for the development of personalized medicine. We have developed a method for SNP prediction based on support vector machines (SVMs) using different features extracted from the SNP data. Prediction rates of 60.9% was achieved by sequence feature, 59.1% by free-energy feature, 58.1% by GC content feature, 58.0% by melting temperature feature, 56.2% by enthalpy feature, 55.1% by entropy feature and 54.3% by the gene, exon and intron feature. We introduced a new feature, the SNP distribution score that achieved a prediction rate of 77.3%. Thus, the proposed SNP prediction algorithm can be used to in SNP discovery.

Algorithms↗

PubMed2007

Identification of differentially expressed genes in multiple microarray experiments using discrete fourier transform.

Research in the post-genome sequence era has been shifting towards a functional understanding of the roles and relationships between different genes in different conditions. While the advances in genetic expression profiling techniques including microarrays enable detailed and genome-scale measurements, the extraction of meaningful information from large datasets remains a challenging task. Here, we propose a novel method of generating gene differential expression profiles such that gene expression values from one dataset can be directly compared with those of another dataset. A simplified Discrete Fourier Transform is applied to interposed gene expression values, thereby generating the 'spectra' for a pair of conditions. Using this technique, differentially expressed genes produce higher amplitudes at the Nyquist Frequency. By measuring the phase of the 'spectra' generated, the over- and under-expressed nature of the genes can be identified. This method was validated using two sets of GeneChip array data, one from prostate cancer related dataset and the other from macular degeneration related dataset. The genes identified as differentially expressed by our method were found to be similar to those published using their preferred methods. Based on our findings, the proposed DFT method could be used efficiently in identifying differentially expressed genes from multiple-array experiments from two different conditions.

Fourier Analysis↗