Normalization

Overview

Phaedra offers a set of normalization methods that turn raw well feature values into normalized feature values. Normalized values are relative to the plate’s high and/or low control wells, and thus make it easier to compare values across plates.

Note: normalization is always performed against control wells within the same plate. Also note that low and high controls do not necessarily map to LC and HC wells. You can map your own control types in the protocol:

control_type_mapping.png

In the below equations, the following names are used:

  • raw: The raw feature value
  • normalized: The normalized feature value
  • lowMedian: The median of the raw values of the low control wells
  • highMedian: The median of the raw values of the high control wells
  • lowStdDev: The standard deviation of the raw values of the low control wells
  • highStdDev: The standard deviation of the raw values of the high control wells

%Effect

$$ normalized = \frac{raw-lowMedian}{highMedian-lowMedian} \times 100 $$

%Effect (100-x)

$$ normalized = 100-\frac{raw-lowMedian}{highMedian-lowMedian} \times 100 $$

%Effect (-x)

$$ normalized = \frac{lowMedian-raw}{highMedian-lowMedian} \times 100 $$

%Effect (x-100)

$$ normalized = \frac{raw-lowMedian}{highMedian-lowMedian} \times 100-100 $$

Sigma [H]

$$ normalized = \frac{raw-highMedian}{highStdDev} $$

Sigma [L]

$$ normalized = \frac{raw-lowMedian}{lowStdDev} $$

%CTLMin

This normalization method is identical to %Effect

%CTL

$$ normalized = \frac{raw}{highMedian} \times 100 $$

%CTL [L]

$$ normalized = \frac{raw-lowMedian}{lowMedian} \times 100 $$

%CTL [L=0]

This normalization method is identical to %CTL [L]

%CTL [INV L=0]

This normalization method is the inverted value of %CTL [L=0] $$ normalized = -\frac{raw-lowMedian}{lowMedian} \times 100 $$

%CTL [L=100]

$$ normalized = \frac{raw}{lowMedian} \times 100 $$

%CTL [H]

$$ normalized = \frac{highMedian-raw}{highMedian} \times 100 $$