Nearest Neighbor Networks for Clustering Gene Expression Data
Author Information
Author(s): Huttenhower Curtis, Flamholz Avi I, Landis Jessica N, Sahi Sauhard, Myers Chad L, Olszewski Kellen L, Hibbs Matthew A, Siemers Nathan O, Troyanskaya Olga G, Coller Hilary A
Primary Institution: Princeton University
Hypothesis
Can the Nearest Neighbor Networks algorithm effectively cluster genes based on their expression profiles?
Conclusion
The Nearest Neighbor Networks algorithm is a valuable clustering method that effectively groups genes that are likely to be functionally related.
Supporting Evidence
- NNN produced clusters with high precision and a broader selection of biological processes than other methods.
- NNN was evaluated on six different microarray datasets, demonstrating its effectiveness across various biological conditions.
- NNN allows genes with no sufficiently similar partners to remain unclustered, improving the quality of the clusters.
Takeaway
This study introduces a new way to group genes that work together by looking at their closest neighbors, helping scientists understand how genes interact.
Methodology
The Nearest Neighbor Networks algorithm uses a graph-based approach to cluster genes based on mutual nearest neighborhoods.
Potential Biases
NNN's performance can be affected by the presence of highly correlated genes, such as ribosomal genes.
Limitations
NNN may not perform as well on datasets with high condition counts or unordered data.
Participant Demographics
The study utilized six different Saccharomyces cerevisiae microarray datasets.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website