McAlister G, Bailey D, Huguet R, and Senko M.
ASMS 2016 Poster
We demonstrate a method of spectral annotation that is based around logistic regression models. The logistic models are trained using isotopic envelopes with known charge state and monoisotopic m/z assignments. With the trained models, we can rapidly identify unknown isotopic envelopes using a simple matrix multiplication step. With this workflow, we identify twice as many peaks as we would have identified with a THRASH-based algorithm3. There is a modest increase in the peak false-annotation rate with the new algorithm, and the average spectral annotation time is ~60 ms.
Thermo Fisher Scientific