Table 1.

Summary of characteristics of methods included in the performance evaluations and runtime comparisons

Method	Spatial information	Flexible length scale parameters	Covariates for spatial domains	Runtime
nnSVG	●	●	●	◐
SPARK-X	●	○	●	●
HVGs	○	○	○	●
Moran’s I	●	○	○	◐
SpatialDE	●	●	●	○
SPARK	●	●	●	○

For each method, the columns indicate whether (●) or not (○) the method: (i) takes spatial information into account, (ii) fits models with flexible gene-specific length scale parameters, (iii) provides an option to include covariates for spatial domains in the models, and (iv) provides fast runtimes. Half-filled circles (◐) indicate intermediate scores. The scalable methods are shown in the first 4 rows, and the earlier cubically scaling methods (SpatialDE and SPARK) are shown in the last 2 rows.