PeptDeepLibraryPredictor Class Reference

Convert peptide precursors into an OpenSWATH assay batch using native PeptDeep ONNX models. More...

#include <OpenMS/ANALYSIS/OPENSWATH/PeptDeepLibraryPredictor.h>

Classes
struct	Config

Public Member Functions
	PeptDeepLibraryPredictor ()
	Construct with the installed OpenMS PeptDeep models and default inference settings.
	PeptDeepLibraryPredictor (const Config &config)
	Construct with explicit inference/model configuration.
OpenSwath::LightTargetedExperiment	predict (const std::vector< PeptDeepLibraryPrecursor > &precursors)
	Predict one batch of precursor assays.

Private Attributes
PeptDeepRTInference	rt_predictor_
std::unique_ptr< PeptDeepCCSInference >	ccs_predictor_
PeptDeepMS2Inference	ms2_predictor_

Detailed Description

Convert peptide precursors into an OpenSWATH assay batch using native PeptDeep ONNX models.

Predicts retention time and MS2 fragment intensities for every precursor. CCS prediction is optional; when enabled, predicted CCS values are converted to reduced inverse mobility (1/K0) through IMTypes and stored as the compound drift time and transition precursor IM.

MS2 output is mapped using the AlphaPeptDeep channel order [b_z1,b_z2,y_z1,y_z2,b_modloss_z1,b_modloss_z2,y_modloss_z1,y_modloss_z2]. Only the four canonical non-loss channels are emitted in this initial implementation.

This class deliberately does not perform digestion, assay filtering, decoy generation, or OpenSWATH ID normalization. Those remain responsibilities of the surrounding library workflow.

Available in builds with WITH_ONNX.

---

Class Documentation

OpenMS::PeptDeepLibraryPredictor::Config

struct OpenMS::PeptDeepLibraryPredictor::Config

Class Members
Size	batch_size = 500
string	ccs_model_path
int	intra_op_threads = 4	threads of each predictor: its model runs that run at once, each on one thread (ONNXPredictorBase::forEachBatch())
string	ms2_model_path
bool	predict_ccs = true
string	rt_model_path

Constructor & Destructor Documentation

PeptDeepLibraryPredictor() [1/2]

PeptDeepLibraryPredictor

(

)

Construct with the installed OpenMS PeptDeep models and default inference settings.

PeptDeepLibraryPredictor() [2/2]

PeptDeepLibraryPredictor ( const Config &  config )

explicit

Construct with explicit inference/model configuration.

Member Function Documentation

predict()

OpenSwath::LightTargetedExperiment predict

(

const std::vector< PeptDeepLibraryPrecursor > &

precursors

)

Predict one batch of precursor assays.

Parameters

precursors

Precursors with caller-assigned unique IDs.

Returns

A LightTargetedExperiment containing one compound per precursor and four b/y fragment channels per cleavage position.

Exceptions

Exception::IllegalArgument	for empty input, empty/duplicate IDs, non-positive precursor charges, peptide sequences shorter than two residues, or empty protein IDs.
Exception::InvalidValue	if an inference output has an unexpected shape.

Member Data Documentation

ccs_predictor_

std::unique_ptr<PeptDeepCCSInference> ccs_predictor_

private

ms2_predictor_

PeptDeepMS2Inference ms2_predictor_

private

rt_predictor_

PeptDeepRTInference rt_predictor_

private