Label-free, ID-Free, Parameter-Free Peak Picking and Feature Selection For LC-MS Data

摘要

Compared to hand labeling: (1) Much faster (2) Captures features not visible in TOPPView (too small) Compared to other methods: (1) No parameters (2) Fully automated-no user in the loop (3) Won't run out of memory (max 2GB ram required) (4) Decisions based on statistics in data, allowing each feature to have different Gaussian parameters (unlike Matched Filtering) (5) Can extract both XIC and spectral features (6) Runs quickly Future work: (1) More examples (different instruments, resolutions, samples) (2) Comparison with other automatic methods (3) Implement in C++ for speed (currently written in Ruby)