UCLA

Dimensionality reduction is nearly ubiquitous in the analysis of single cell sequencing data. However, until the current work, no serious effort had been made to quantify the distortion introduced by dimensionality reduction and the effect of that distortion on the analysis. Here, I first present a method for the measurement of distortion caused by dimensionality reduction, Average Jaccard Distance. I will show that the application of this metric to data analysis workflows suggests the need for revision in the way that these methods are used for single cell RNA sequencing analysis. Next, I propose a revised methodology, and present the results of applying this revised methodology to the study of small cell lung cancer. The results include the identification of a stem-like population of cancer cells and many potential drug targets. Finally, I present the schematic of a new, more accurate method of dimensionality reduction using deep neural networks.