Dimensionality reduction & clustering / t-SNE
Your data stays in your browser: files are read and processed on your device, never uploaded to any server, and cleared when you close the page. Free, no sign-in.
t-SNE places samples that are close in high-dimensional space close together in the plane, which shows whether cells or samples form clusters; it mainly preserves local neighbourhoods.
Cluster sizes, distances between clusters and the coordinate values themselves are not comparable, so do not read “how different” two groups are from the plot; use the original data or PCA for that.
Perplexity is roughly the effective number of neighbours per point: 5–50 is typical, smaller for few samples. Different perplexities or seeds give differently shaped plots; with a fixed seed the result is reproducible.
For gene-expression matrices log-transform first and reduce to the top 30–50 principal components (default 50). Trustworthiness (0–1, higher is better) measures how well local neighbours are preserved.
This is exact t-SNE (O(n²)); in a browser 1,000 samples take from about ten seconds to a minute, so set the parameters and then press “Run”.
van der Maaten L. & Hinton G. (2008) J Mach Learn Res 9:2579–2605. Parameters and optimisation follow scikit-learn TSNE(method="exact"); because the random sequences differ, coordinates are not identical point by point, but on the example data trustworthiness and KL divergence agree within 0.02 and 10%.
Up to 5,000 samples.
Send us your data and research question and you will receive a written plan within 1 working day: analysis steps, parameter rationale, deliverables and timeline. Quoted per project.