Novel Algorithm for Analyzing Whole Genome Amplified DNA for SNP Array Analysis
MA Sajid, X Zhou, W-T Huang, C-C Chang. The Methodist Hospital, Houston, TX; The Methodist Hospital Research Institute, Houston, TX
Background: Applying whole genome amplification to some fractions is required to generate enough DNA material for single nucleotide polymorphism (SNP) array analysis using the sorting approach. Using whole genome amplification, noise is amplified decreasing the signal noise ratio (SNR) of log2-ratio values.
Design: To reduce the influence of noise, we propose a conditional random pattern (CRP) model for copy number inference using amplified DNA samples. In CRP model, the current hidden state (real copy number) is determined by its two immediate neighboring hidden states, and a continuous segment of observations (log2-ratios). We evaluated the feasibility of using the amplified DNA for SNP array analysis by Pearson correlation of log2-ratio values between whole-genome amplified DNA and un-amplified DNA. The calculated Pearson correlation coefficient between un-amplified and amplified DNA samples is about 0.81. We then applied dChip, CNAG and CRP model to 6 SNP arrays of myelodysplastic syndrome (MDS) samples after whole-genome amplification. dChip and CNAG are academic softwares for SNP array analysis. Copy number inference results using dChip, CNAG and CRP model in a case with trisomy (Chromosome 8) and a case with monosomy (Chromosome 7), as determined by conventional cytogenetic studies and by SNP array analysis using non-whole-genome amplified DNA of the same patient, were compared.
Results: The results showed that dChip and CNAG, but not CRP, miss many gain regions in chromosome 8 and deletion regions in chromosome 7. Additionally, dChip and CNAG detected about 100 copy number aberration (CNA) regions in other chromosomes in both cases, with 19 of them consisting of only one SNP locus. Most of these changes are not observed when the data are analyzed by CRP or when the SNP array using non-amplified DNA of the same patients were analyzed by dChip, CNAG or CRP.
Conclusions: These results indicated that the CRP model we developed generates more accurate results using amplified DNA samples than dChip or CNAG.
Tuesday, March 10, 2009 11:15 AM
Platform Session: Section G2, Tuesday Morning