Comparing Neural Networks and Other Methods for Genetic Association Studies
Author Information
Author(s): Curtis David
Primary Institution: Academic Centre for Psychiatry, St Bartholomew's and Royal London School of Medicine and Dentistry
Hypothesis
Is artificial neural network analysis more effective than traditional multimarker methods for detecting genetic associations?
Conclusion
Artificial neural network analysis is more powerful than standard haplotype-based tests, but its complexity and slow performance may limit its adoption.
Supporting Evidence
- ANN analysis had the highest power compared to other methods.
- The difference in power between ANN and haplotype-based tests was statistically significant.
- Permutation testing was more powerful than asymptotic tests for haplotype analysis.
- The t statistic from ANN analysis correlated highly with empirical p values.
- Sample sizes and disease models significantly influenced the ability to detect associations.
Takeaway
This study looked at different ways to find links between genes and diseases. It found that a special computer method called neural networks works better than older methods, but it's slower and harder to use.
Methodology
The study compared the power of various genetic analysis methods using simulated SNP datasets.
Limitations
The ANN method is slow and lacks a theoretical model for its results.
Participant Demographics
60 unrelated subjects from the HAPMAP project.
Statistical Information
P-Value
0.001
Statistical Significance
p = 0.001
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website