Weird UMAP plot - Index Column in Data

[plath997]

Hi, I am trying to run UMAP on my dataset (I am using R via the reticulate package to run umap on a Seurat object). However, my dataset has yielded very odd results.

My dataset has 23,000 cells and 12,000 genes. These are the parameters that I have been using:

UMAP(a=None, angular_rp_forest=False, b=None, init='spectral',
     learning_rate=1.0, local_connectivity=1, metric='correlation',
     metric_kwds=None, min_dist=0.5, n_components=2, n_epochs=None,
     n_neighbors=30, negative_sample_rate=5, random_state=None,
     repulsion_strength=1.0, set_op_mix_ratio=1.0, spread=1.0,
     target_metric='categorical', target_metric_kwds=None,
     target_n_neighbors=-1, target_weight=0.5, transform_queue_size=4.0,
     transform_seed=42, verbose=True)

I have tried adjusting the min_dist a and b to no avail. Any advice would be greatly appreciated.

[lmcinnes]

That looks like the sort of results you can get when an index column made it into the data by accident (I have certainly done this myself once or twice). You may want to recheck the data you have and whether there are any weird feature columns that are actually an ordered index.

[josh-gree]

Lets say I am certain that I don't have an index column in my data - if we see a UMAP that looks like this what can we say about the data that went in? Would it be correct to say that if we see these "worm" like artefacts then our data lives in some 1d subspace?

I think the ID thing is probably essentially saying the same thing right? Some component of the input lives on a line?

[jlmelville]

Don't know if this is helpful, but you can see a similar structure when embedding a series of random walks. https://github.com/jlmelville/drnb/blob/master/notebooks/data-pipeline/randomwalks.ipynb