Additional file 3: Figure S3. of SLICER: inferring branched, nonlinear cellular trajectories from single cell RNA-seq data Joshua Welch Alexander Hartemink Jan Prins 10.6084/m9.figshare.c.3615848_D9.v1 https://springernature.figshare.com/articles/journal_contribution/Additional_file_3_Figure_S3_of_SLICER_inferring_branched_nonlinear_cellular_trajectories_from_single_cell_RNA-seq_data/4377824 Additional simulations comparing SLICER with other approaches. (a) The first three functions used to generate the synthetic data discussed in Fig. 2. Note the highly curved shape of the trajectory. (b) The first three functions used to generate an additional dataset. This trajectory is much less curved than the one shown in panel (a), and ICA thus performs much better on this example. (c) Performance of SLICER and other approaches, with and without gene selection, on the trajectory shown in panel (b) as the proportion of irrelevant genes increases. Note that the other approaches do not perform gene selection on their own, so the genes selected by SLICER were given as input for this comparison. A noise level of 2 was used for these simulations. Note that the y-axis does not start at 0. (d) Performance of SLICER and other approaches on the trajectory shown in panel (b) as the noise level increases. To isolate the effect of increasing noise, an irrelevant gene proportion of p = 0 was used for these datasets. Note that the y-axis does not start at 0. (PDF 146 kb) 2016-05-23 05:00:00 Single cell RNA-seq Time series Manifold learning