MaskedPatterns negation maybe reversed

[trvrb]

@rambaut ---

Here are initial results for the 4 way rate split between:

• epitope trunk
• nonepitope trunk
• epitope branch
• nonepitope branch

This was from running H3N2_stem_partition.xml:

This appears backwards compared to the results of Wolf et al. 2006:

They find epitopes are much faster than nonepitopes by almost 10x ratio. We have a ratio of ~3X in favor of nonepitope sites. Within epitope sites they find the trunk faster than side branches. In nonepitope sites, we find the trunk faster than side branches.

Here is the bit of code from MaskedPatternParser:

for (char c : maskString.toCharArray()) {
    if (Character.isDigit(c)) {
        if (k >= mask.length) {
            break;
        }
        mask[k] = (c == '0' ? negativeMask : !negativeMask);
        k++;
    }
}

Does it look like negativeMask could be reversed?

[rambaut]

The code looks correct to me. If the maskString has character 0 then the site is included if negativeMask is true, not otherwise. If the maskString has character 1 then it will be included if the negativeMark is false.

[trvrb]

Yeah. Sorry I just noticed this. I had: <maskedPatterns id="epitopePatterns" negative="true"> and <maskedPatterns id="nonepitopePatterns" negative="false">. It was my XML.

Good to know things are working (more or less --- trends at the moment weaker than what Wolf et al. found).