Fig. 2: The built-in and pre-trained MS2, RT, and CCS prediction models.

The MS2 model is built on four transformer layers, and the RT/CCS models consist of a convolutional neural network (CNN) layer followed by two bidirectional long short-term memory (BiLSTM) layers. The pre-trained MS2 model currently supports predicting the intensities of backbone b/y ions as well as their modification-associated neutral losses if any (e.g. –98 Da loss of phosphorylation on Ser/Thr). However, the user can easily configure the MS2 model to train and predict water and ammonium losses from backbone fragments as well. RT retention time, CCS collision cross section, MS2 intensities of fragment spectra, BiLSTM bidirectional long short-term memory, CNN convolutional neural network, AA amino acid, PTM post-translational modification, NCE normalized collision energy.