Biology is being inundated by an onslaught of noisy, high-dimensional data to an extent never before experienced

Dimensionality reduction techniques such as principal component analysis (PCA) are common approaches for dealing with noisy, high-dimensional data. Though these unsupervised techniques can help uncover interesting structure in high-dimensional data they give little insight into the biological and technical considerations that might explain the uncovered structure.

Here researchers from MIT introduce a hybrid approach-component selection using mutual information (CSUMI)-that uses a mutual information-based statistic to reinterpret the results of PCA in a biologically meaningful way. They apply CSUMI to RNA-seq data from GTEx. This hybrid approach enables them to unveil the previously hidden relationship between principal components (PCs) and the underlying biological and technical sources of variation across samples. In particular, the researchers look at how tissue type affects PCs beyond the first two, allowing us to devise a principled way of choosing which PCs to consider when exploring the data. They further apply their method to RNA-seq data taken from the brain and show that some of the most biologically informative PCs are higher-dimensional PCs; for instance, PC 5 can differentiate the basal ganglia from other tissues. They also use CSUMI to explore how technical artifacts affect the global structure of the data, validating previous results and demonstrating how this method can be viewed as a verification framework for detecting undiscovered biases in emerging technologies. Finally the researchers compare CSUMI to two correlation-based approaches, showing theirs outperforms both.


(a) A plot of the RNA-seq data projected onto the first two PCs, where each spot corresponds to a sample and each color to a tissue type of origin. We see that, as expected, the samples from the same tissue cluster together. (b) A plot of the RNA-seq data projected onto the fifth and seventh PCs, now colored by enrollment center (i.e., BSS center codes C1, B1, and D1). We see that there is an obvious relation between the seventh PC and enrollment center. PCs, principal components.

Availability – A python implementation is available online on the CSUMI website:

Simmons S, Peng J, Bienkowska J, Berger B. (2015) Discovering What Dimensionality Reduction Really Tells Us About RNA-Seq Data. J Comput Biol [Epub ahead of print]. [article]

Leave a Reply

Your email address will not be published. Required fields are marked *


Time limit is exhausted. Please reload CAPTCHA.