Abstract
Accurate and comprehensive extraction of information from high-dimensional single cell datasets necessitates faithful visualizations to assess biological populations. A state-of-the-art algorithm for non-linear dimension reduction, t-SNE, requires multiple heuristics and fails to produce clear representations of datasets when millions of cells are projected. We developed opt-SNE, an automated toolkit for t-SNE parameter selection that utilizes Kullback-Liebler divergence evaluation in real time to tailor the early exaggeration and overall number of gradient descent iterations in a dataset-specific manner. The precise calibration of early exaggeration together with opt-SNE adjustment of gradient descent learning rate dramatically improves computation time and enables high-quality visualization of large cytometry and transcriptomics datasets, overcoming limitations of analysis tools with hard-coded parameters that often produce poorly resolved or misleading maps of fluorescent and mass cytometry data. In summary, opt-SNE enables superior data resolution in t-SNE space and thereby more accurate data interpretation.
Footnotes
Multiple sections of the manuscript were updated and text revised. Specifically, Figure 5 was added to provide computation time benchmarks, accuracy scores, and scalability data of opt-SNE and corresponding text in Results and Discussion sections was revised/updated; supplemental figures were updated and expanded. Data and code availability section was updated to include links to publicly available code (C++ with a Python wrapper) with instructions on installation/usage, a cloud version of opt-SNE and datasets mentioned in the paper (http://www.omiq.ai/opt-SNE). Text of the manuscript was revised for better clarity and the abstract was re-written.
