$ q_{com}$

The composite quality score is a combination of several of the quality measured above, and is is given by

$\displaystyle q_{com}=(q_{size} \cdot q_{sig-noise} \cdot q_{bg1} \cdot
 q_{bg2})^{1/4} \cdot I_{sat},$    

where $ I_{sat}$ is an indicator function depending on if the particular spot is defined as saturated or not, i.e.

$\displaystyle I_{sat}=\left\{ \begin{array}{ll}
 1, & \mbox{if the number of sa...
... the number of saturated pixels $ \geq \;\; 10(50)\% $.} 
 \end{array}
 \right.$    

According to [5], this gives a good measurement of data variability, and can therefore be used to improve data reliability. There are several possibilities for defining this score, and often the choice depends on the problem at hand. Other choices are listed in [4] and [5].

In most articles on this subject, a cut-off value is chosen for the various quality measures, and spots not satisfying these conditions are removed from further analysis. According to [1], one should be careful when doing this though, since sometimes what looks like a problem with the data, may come from natural causes, thus removing them may introduce bias, instead of removing it. Weighting spots, depending on their quality score, instead of removing them, could therefore be a better choice. This allows you to continue to use spots when doing e.g. normalisation. Another reason for introducing weights instead of cut-off values is that, according to [2], the values on the quality scores are very dependent on the methods used during the image analysis.

Bjørn Kåre Alsberg 2006-04-06