How to include 3+ fragments?

[cctsou]

Hi,

We are trying to identify longer petptides and therefore higher charge states and 3+ fragments become important to us. I have a few DDA library with 5+, 6+ peptide ions included also with all 1+, 2+, and 3+ fragments. I changed the yaml file

max_frag_charge: 3

and I was able to perform transfer learning without problem, but then after I generate predicted library from FASTA file, still only 1+, 2+ fragments were included in the tsv file.

Could you please advise me on how to include 3+ fragments? Here is my yaml file:

model:
  frag_types:
  - b
  - y
  - b_modloss
  - y_modloss
  max_frag_charge: 3
PEPTDEEP_HOME: C:\Users\Administrator/peptdeep
local_model_zip_name: pretrained_models.zip
model_url: https://github.com/MannLabs/alphapeptdeep/releases/download/pre-trained-models/pretrained_models.zip
task_workflow:
- library
task_choices:
- train
- library
thread_num: 40
MAX_THREADS: 60
torch_device:
  device_type: gpu
  device_type_choices:
  - get_available
  - gpu
  - mps
  - cpu
  device_ids: []
log_level: info
log_level_choices:
- debug
- info
- warning
- error
- critical
common:
  modloss_importance_level: 1.0
  user_defined_modifications:
    IADTB@C:
      composition: H(24)C(14)N(4)O(3)
      modloss_composition: ''
peak_matching:
  ms2_ppm: true
  ms2_tol_value: 20.0
  ms1_ppm: true
  ms1_tol_value: 20.0
model_mgr:
  default_nce: 30.0
  default_instrument: Lumos
  mask_modloss: true
  model_type: generic
  model_choices:
  - generic
  - phos
  - hla
  - digly
  external_ms2_model: C:/Users/Administrator/peptdeep/refined_models_v3/ms2.pth
  external_rt_model: C:/Users/Administrator/peptdeep/refined_models_v3/rt.pth
  external_ccs_model: C:/Users/Administrator/peptdeep/refined_models_v3/ccs.pth
  instrument_group:
    ThermoTOF: ThermoTOF
    Astral: ThermoTOF
    Lumos: Lumos
    QE: QE
    timsTOF: timsTOF
    SciexTOF: SciexTOF
    Fusion: Lumos
    Eclipse: Lumos
    Velos: Lumos
    Elite: Lumos
    OrbitrapTribrid: Lumos
    ThermoTribrid: Lumos
    QE+: QE
    QEHF: QE
    QEHFX: QE
    Exploris: QE
    Exploris480: QE
    THERMOTOF: ThermoTOF
    ASTRAL: ThermoTOF
    LUMOS: Lumos
    TIMSTOF: timsTOF
    SCIEXTOF: SciexTOF
    FUSION: Lumos
    ECLIPSE: Lumos
    VELOS: Lumos
    ELITE: Lumos
    ORBITRAPTRIBRID: Lumos
    THERMOTRIBRID: Lumos
    EXPLORIS: QE
    EXPLORIS480: QE
  predict:
    batch_size_ms2: 512
    batch_size_rt_ccs: 1024
    verbose: true
    multiprocessing: true
  transfer:
    model_output_folder: C:/Users/Administrator/peptdeep/refined_models
    epoch_ms2: 20
    warmup_epoch_ms2: 10
    batch_size_ms2: 512
    lr_ms2: 0.0001
    epoch_rt_ccs: 40
    warmup_epoch_rt_ccs: 10
    batch_size_rt_ccs: 1024
    lr_rt_ccs: 0.0001
    verbose: false
    grid_nce_search: false
    grid_nce_first: 15.0
    grid_nce_last: 45.0
    grid_nce_step: 3.0
    grid_instrument:
    - Lumos
    psm_type: alphapept
    psm_type_choices:
    - alphapept
    - pfind
    - maxquant
    - diann
    - speclib_tsv
    - msfragger_pepxml
    - spectronaut_report
    dda_psm_types:
    - alphapept
    - pfind
    - maxquant
    - msfragger_pepxml
    psm_files: []
    ms_file_type: alphapept_hdf
    ms_file_type_choices:
    - alphapept_hdf
    - thermo_raw
    - mgf
    - mzml
    ms_files: []
    psm_num_to_train_ms2: 100000000
    psm_num_per_mod_to_train_ms2: 50
    psm_num_to_test_ms2: 0
    psm_num_to_train_rt_ccs: 100000000
    psm_num_per_mod_to_train_rt_ccs: 50
    psm_num_to_test_rt_ccs: 0
    top_n_mods_to_train: 10
    psm_modification_mapping: {}
library:
  infile_type: fasta
  infile_type_choices:
  - fasta
  - sequence_table
  - peptide_table
  - precursor_table
  - all_other_psm_reader_types
  infiles:
  - C:/Users/Administrator/peptdeep/uniprot_swissprot_20200903_can_iso_with_LOH_altAlleles_Ver3_20230808_fragpipe.fasta
  fasta:
    protease: ([DE])
    protease_choices:
    - trypsin
    - ([KR])
    - trypsin_not_P
    - ([KR](?=[^P]))
    - lys-c
    - K
    - lys-n
    - \w(?=K)
    - chymotrypsin
    - asp-n
    - glu-c
    max_miss_cleave: 5
    add_contaminants: false
  fix_mods:
  - IADTB@C
  var_mods:
  - Acetyl@Protein_N-term
  - Oxidation@M
  special_mods: []
  special_mods_cannot_modify_pep_n_term: false
  special_mods_cannot_modify_pep_c_term: false
  labeling_channels: {}
  min_var_mod_num: 0
  max_var_mod_num: 2
  min_special_mod_num: 0
  max_special_mod_num: 1
  min_precursor_charge: 2
  max_precursor_charge: 6
  min_peptide_len: 5
  max_peptide_len: 50
  min_precursor_mz: 200.0
  max_precursor_mz: 2000.0
  decoy: None
  decoy_choices:
  - protein_reverse
  - pseudo_reverse
  - diann
  - None
  max_frag_charge: 3
  frag_types:
  - b
  - y
  rt_to_irt: false
  irt_library: xxx/library.tsv
  irt_library_type: speclib_tsv
  generate_precursor_isotope: false
  output_folder: C:/Users/Administrator/peptdeep/spec_libs_uniprot_swissprot_20200903_can_iso_with_LOH_altAlleles_Ver3_20230808_fragpipe_gluc_v3_nodecoy
  output_tsv:
    enabled: true
    min_fragment_mz: 200.0
    max_fragment_mz: 2000.0
    min_relative_intensity: 0.001
    keep_higest_k_peaks: 12
    translate_batch_size: 100000
    translate_mod_to_unimod_id: false

[jalew188]

This requires to create a new model and train it

[cctsou]

This requires to create a new model and train it

Could you guide me on how to do that with PeptDeep? Just leave pretrained model field empty and still use transfer command?

[cctsou]

@jalew188 ,

I tried retraining a model by leaving "pretrained model field" empty and using the transfer command, but I am still not getting 3+ fragment in the library I generated from FASTA file. Could you advise any way I could get it resolved?

[jalew188]

No, I mean you need to write new model code, then train the new model. This is not enabled by command line.