Mode shape robustness to number of points used to parameterise the outline, and independent components analysis.
(a) 500 points: shapes and variance explained are the same as for Figure 1 with 250 points (five significant components). Using 1,000 points also produced identical results (data not shown). (b) 100 points: shapes are the same, but variance explained is slightly different (four significant components). (c) 50 points: both shapes and variance explained are now noticeably different (three significant components). (d) 25 points: shapes and variance explained are now very different (two significant components). Thus 250 points is a sufficiently detailed parameterisation such that no further changes in shape or number of significant components occur as this number is increased. (e) 20 independent components were found for the in vitro (no gradient) dataset using the FastICA algorithm . Ordering is arbitrary. Some of these components (e.g., M5 to M7) appear to be trying to represent individual filopodia.