Identifying SNPs Associated with Rheumatoid Arthritis
Author Information
Author(s): Meng Yan, Yang Qiong, Cuenco Karen T, Cupples L Adrienne, DeStefano Anita L, Lunetta Kathryn L
Primary Institution: Boston University
Hypothesis
Can a two-stage approach using random forests and Bayesian networks effectively identify SNPs associated with rheumatoid arthritis?
Conclusion
The study successfully filtered out many unassociated SNPs while retaining important risk SNPs for rheumatoid arthritis.
Supporting Evidence
- The random forest analysis reduced the number of variables for further Bayesian network analyses from 9190 to about 53.
- 94 out of 100 replicates retained important SNPs associated with rheumatoid arthritis.
- The Bayesian network analysis further decreased the test set prediction error.
- CaMML identified all covariates and almost all surrogates in linkage disequilibrium with disease loci.
Takeaway
Researchers used a two-step method to find genetic markers linked to rheumatoid arthritis, which helps in understanding the disease better.
Methodology
The study used random forests to screen SNPs and Bayesian networks to analyze interactions among selected variables.
Potential Biases
The iterative method may introduce bias due to overfitting.
Limitations
The study may not generalize to data with weaker association signals and has potential overfitting issues.
Participant Demographics
The study involved 1500 nuclear families with affected siblings and 2000 unrelated unaffected individuals as controls.
Statistical Information
P-Value
5.26 × 10-18
Statistical Significance
p<0.0001
Want to read the original?
Access the complete publication on the publisher's website