Start from raw data?

PyroNoise (see discussion under ‘Journal Club’) suggests that 454 pyrosequencing data quality could be improved by enhanced base calling from the raw data. Could this be an option for incorporation into mothur? I notice, for example, that PyroNoise does quality screening based on raw flowgram data, while mothur does quality screening from the (derived) quality scores. Does anyone know what the implications might be, or how flowgram signals are converted to quality scores?

I’m on it. The next version of mothur will have something that allows people to work with sff files. It may be a bit immature of a feature, but I’m anxious to get people a more user-friendly way to use pyronoise. Based on some unpublished studies that I’ve seen as posters, it looks like the qual scores are next to meaningless.