Fig. 7

The plot shows runtimes (s) for the imputation algorithms under consideration. DrImpute and scImpute did not scale favorably with data size and timed out at the maximum allotted time of 160 minutes on the simulated datasets. MAGIC was the fastest method with polynomial runtime complexity. ccImpute was the second fastest on datasets with less than 8000 cells, with DeepImpute taking the second fastest after 8000 cells. This shows ccImpute has competitive runtime, and with further optimizations such as swapping K-means for a faster clustering algorithm, this approach may take the lead in both runtime performance and quality of imputation