Pathogen Chip Sensitivity

CREATED: 200701051243 ** Pathogen Chip

sensitivity not as high as real time PCR (traditional method) due to cross hybridization
microarray with 40-mer problems, tiling array
microarray is based on whole genome whereas real time PCR only selects a specific portion of the virus genome may fail to detect if there are mutations in that region

** Decting unknown virus

** Random PCR

primer consists of fix head - variable tail
tail portion consists of all possible letters to match different parts of virus RNA
problems - inefficient amplification ** need both forward and reverse primer to bind ** distance from 500 to 1000 bp for PCR to work
factors which affect binding ** temperature, different primer require different temperature as binding strength of A-T != C-G ** intra primer secondary structure
solution - build model to estimate probability of forward/backward primer binding at site i
base on this probability, compute an Amplification Efficiency Score (AES)
AES correlates well with hybridization signal intensity

** Application of AES

predict signal intensity of different regions
find better random primers ** head portion of primer also affects binding ** select head that have high AES for more regions

** Detection of pathogen

t-statistic of average signal ** high false positive ** some false negative
Observation: real virus has heavy tail ** use KL divergence to check difference in shape ** put more importance in tail portion -> weighted KL (WKL)

** Reference