Scan-Wise Activation and Peak Selection

Introduction

Scan-Wise Activation and Peak Selection (SWAPS) is a novel modular MS1-centric PIP framework designed to elevate peptide identification and quantification by leveraging the utilization of MS1-data at all available dimensions, state-of-the-art peptide-prediction models, and an innovative deep-learning-based method for feature quality control (peak selection and confidence scoring)

Usage

Environment Setup

SWAPS is developed with Python 3.10 with the environment specificied in environment_cross_platform.yml (cross-platform) or environment_explicit.txt.

conda env create -n SWAPS --file environment_cross_platform.yml

Run SWAPS

Configuration

SWAPS takes a .yaml as a configuration argument. Examples can be found at utils/exp_configs. Please modify the file path as necessary. Detail documents of config definiation can be found at utils/singleton_swaps_optimization.py.

Runner

To run SWAPS, in the repo directory, use:

python sbs_runner_ims.py [path-to-config-file]

SWAPS results structure

• [RESULT_PATH]/

  • config_[TIMESTAMP].yaml (config file copy for reproducibility)

  • ms1scans.csv

  • mobility_values.csv

  • contruct_dict/ (intermediate results during dictionary construction)

    • RT_transfer_learn/
      • ...
    • IM_transfer_learn/
      • ...
    • BarChart_candidate_overlap.png
    • BarChart_exp_elution_counts.png
    • ...

  • maxquant_result_ref.pkl (SWAPS dictionary)

  • peak_selection/ (results from peak selection, only exists if peak_selection is enabled)

    • training_data/
      • ...(sparse matrix (for hint channel) and annotated data from MS2 identification)
    • exp_[TIMESTAMP]/
      • updated_peak_selection_config.yaml
      • logs_tensorflow/
        • ... (for tensorboard visualization)
      • model_backups/
        • ... (weights of segmentation and scoring models)
      • results/
        • evaluation/
          • ... (result on full dataset)
        • ... (other evaluation results on testset)
        • loss.json
        • metric.json
      • pept_act_sum_ps.csv (all candidate results (target+decoy) inferred intensity after peak selection)
      • pept_act_sum_ps_full_fdr_thres.csv (all candidate results (target+decoy) with FDR threshold
      • pept_act_sum_ps_full_tdc_fdr_thres.csv (candidates after target-decoy competition)

  • results/

    • activation/
      • sparse matrix in batch and peptbatch.npz
      • pept_act_sum.csv
      • pept_act_sum_filter_by_im.csv (peptide activation sum after filter by ion mobility, only exists if __C.RESULT_ANALYSIS.POST_PROCESSING.FILTER_BY_IM==True)
    • evaluation/ (only exists if peak selection is disabled, evaluation compared to reference (MQ))
      • CorrQuant.png
      • VennDiag.png

---

For questions, please contact zixuan.xiao@tum.de