How did they find these position specific scoring matrix scores?

Question

[jbrody11]

My lecture notes in Bioinformatics try to explain how to work out position specific scoring matrix, or position weight matrix. They use the following example in the notes. I understand the formula but I don't see how they could calculate -2.94 for a frequency/base count of 0, as log2(0) isn't a number. Could somebody explain how they might have found -2.94 for the 0 base counts? I am also not sure what pb value they used. I assume fbi means the base count (frequency) Image here of their explanation

Answer 1

[Mensur Dlakic]

Zero counts in sequence alignments is a common occurrence for biological sequences. Small pseudocounts are added to avoid numerical issues - see here. In practical terms this means that each residue will have a non-zero frequency.