Table 3.
Resistant | Sensitive | Overall Sensitive % | ||
Channel | Feature Type | |||
Ch1 | arrangement | 24 | 16 | 58% |
dimension | 16 | 94 | ||
intensity | 32 | 18 | ||
shape | 33 | 27 | ||
texture | 14 | 6 | ||
Ch2 | intensity | 19 | 1 | 5% |
Ch3 | arrangement | 6 | 14 | 47% |
intensity | 36 | 14 | ||
texture | 26 | 33 | ||
Ch4 | arrangement | 10 | 20 | 46% |
dimension | 7 | 3 | ||
intensity | 57 | 33 | ||
shape | 2 | 8 | ||
texture | 32 | 28 |
The dose response of each feature for each drug was fitted by a linear model, using either the well-segmented cell population, or the poorly-segmented cell population. The discrepancy between the well- and poorly- segmented dose responses was calculated for each drug-feature combination, and drug-feature combinations were then binned as either sensitive to segmentation (if the discrepancy was larger than the grand median discrepancy) or resistant to segmentation (if the discrepancy was smaller than the grand median).