Quality Controls

Quality controls

Spotfinder uses three types of quality control parameters to evaluate the quality of spot in each channel and in total goodness. The output MEV file reports flags for each channel, QC score total and for every channels, and P values for each channel.

Spotfinder Flags

There are two types of flags. The flags A, B, and C are assigned to the good spots and differentiate good spots by there size.

A - spot area is larger than 50 pixels;
B - spot area is in the range between 30 and 50 pixels;
C - spot area is smaller than 30 pixels;

The flags S, X, Y, and Z are introduced to the rest of the spots. The next is there meaning:

S - spot is fully or partialy saturated. This is a warning flag , user should evaluate spot and make a decision about it.
U - spot is set by user for not to be processed, i.e. ignored.
X - bad spot; this flag is assigned by QC filter. QC filter dependent flag, i.e. when QC filter is off there are no spots with X flag.
Y - bad spot. Spot for which background is higher then signal. QC filter independent flag. This flag presents even when QC filter is off.
Z -spot is not detected. QC filter independent flag.

Spotfinder QC score

QC score is calculated depending on signal-to-noise ratio for every channel in spot and shape quality for whole spot. Spot shape QC score is defined as ratio of spot area to spot perimeter scaled into the range between 0 and 1. It is the same for both channels in the spot, because the shape of spot is identical in all channels for given spot. Signal-to-noise QC score is calculated as percentage of pixels in a spot with values higher than 2*median(local BKG). The QC score in each channel is calculated as geometric mean of shape and Sgnal-to-noise QC scores for this channel. The total spot QC score is taken as a mean of QC scores in all channels.

P values

P-values are calculated for each channel as result of t-test comparing the spot pixel set and surrounding background pixel set. We apply simple procedure based on mean and standard deviation analysis for both spot and background pixel sets. The lower P-values the better spot. One can say that spot with p-value larger than 0.3-0.35 is not good.