Related Experiment Videos
Evaluation of gel-pad oligonucleotide microarray technology by using artificial neural networks.
Alex Pozhitkov1, Boris Chernov, Gennadiy Yershov
1Civil and Environmental Engineering, University of Washington, Seattle, WA 98195, USA.
Applied and Environmental Microbiology
|December 8, 2005
Summary
Artificial neural networks (ANNs) were used to analyze thermal dissociation profiles from DNA microarrays. ANNs identified ideal and poor quality nucleic acid hybridization data, revealing experimental variability impacts accuracy.
Area of Science:
- Molecular Biology
- Bioinformatics
- Genomics
Background:
- Thermal dissociation analysis of nucleic acids on DNA microarrays can enhance duplex discrimination.
- Previous studies suggest broad stringency conditions improve analysis.
- A gel-pad microarray format was previously described.
Purpose of the Study:
- To evaluate the utility of thermal dissociation analysis with artificial neural networks (ANNs) for DNA microarrays.
- To train ANNs to recognize noisy or low-quality hybridization data.
- To classify nucleic acid dissociation profiles (melts) for improved data quality assessment.
Main Methods:
- Artificial neural networks (ANNs) were trained using 21,044 dissociation profiles.
- Data derived from 186 probes hybridized to RNA from 32 oral microbes.
- Profiles were classified based on characteristics like signal intensity and area under the curve.
Main Results:
- Three melt profile groups were identified: ideal, poor, and difficult to classify.
- Approximately 18% of perfect-match duplexes were misclassified as poor.
- Experimental variability, local background subtraction sensitivity, and long RNA fragments affected results.
Conclusions:
- ANNs can classify hybridization quality in thermal dissociation analysis.
- Inconsistencies in melt profiles were observed even for perfect-match hybrids.
- Methodological factors like background subtraction and fragment length impact data reliability.