Quantselect-implementation by vuductung · Pull Request #706 · MannLabs/alphadia

vuductung · 2025-10-23T10:18:01Z

No description provided.

…ultiple normalization methods (directLFQ, quantselect)

…uding model, training, optimizer, and prediction configurations.

…to simplify configuration for users.

…tion support

…ion and enhance parameter handling. Introduced feature dictionary for input data and improved logging for normalization processes. Updated documentation for parameters and added compatibility for legacy configurations.

…zation. Introduced fixtures for MS2 features and PSM files to enhance testing coverage. Updated logging to reflect quantselect processing. Refactored existing tests to accommodate new feature dictionary input structure.

…or required psm_df input and improving configuration handling. Default settings are now merged with user-provided configurations, and the seed for random determinism is dynamically set based on the configuration.

…on method directly in the intensity filtering step. Updated logging for fragment absence and adjusted input handling for lfq method to use the feature dictionary.

…ity. Improved logging format for normalization processes and adjusted DataFrame handling in unit tests for better clarity and maintainability.

…ith quantselect.

…ibility issues.

…d of 'normalize_lfq' for consistency with recent changes in the quantification process.

…turn a dictionary instead of a tuple, improving clarity in handling fragment data. Update SearchPlanOutput to handle cases where no fragment data is found, enhancing logging for better debugging.

…turn a tuple instead of a dictionary, clarifying the handling of empty data cases. Update SearchPlanOutput to remove redundant checks for fragment data absence, improving code efficiency.

mschwoer

some upfront comments regarding the integration and overall structure

mschwoer · 2025-10-27T08:59:40Z

+  # Normalization method for label-free quantification values
+  # Options: "directlfq", "quantselect", "none" (or false for backwards compatibility)
+  normalization_method: "quantselect"


please set the default to directlfq (to prevent previous behaviour)

also, add ... .normalize_lfq to config.py : TOLERATED_KEYS

then the change is only breaking for people that used normalize_lfq: False (probably few)

the "(or false for backwards compatibility)" can be removed

done, set the default as directlfq. - normalize_lfq was kept in defautl_yaml, so adding to TOLERATED_KEYS not necessary?

mschwoer · 2025-10-27T09:00:00Z

  min_nonnan: 3
-  # Enable normalization of label-free quantification values
-  normalize_lfq: True
+  # Normalization method for label-free quantification values


also, frontend needs to be adapted to use the new key and default

mschwoer · 2025-10-27T09:00:16Z

-        )  # here you can chose wether to log the processed proteins or not
+        # Apply normalization based on the selected method
+        if normalize == "quantselect":
+            if psm_df is None:


please move all code in this if branch to a new class/file

(cf also @GeorgWa 's recent refactorings(

mschwoer · 2025-10-27T09:02:42Z

            "num_samples_quadratic": 50,
            "min_nonnan": 1,
-            "normalize_lfq": True,
+            "normalization_method": "quantselect",


also here, please use directlfq as default

mschwoer · 2025-10-27T09:03:03Z

 - Empirical library and fully predicted library search
 - End-to-end transfer learning for custom RT, mobility, and MS2 models
- Label free quantification
+- Label free quantification with multiple normalization methods (directLFQ, quantselect)


maybe add links to the repective repos?

mschwoer · 2025-10-27T09:03:38Z

        min_nonan: int = 1,
        num_cores: int = 8,
-        normalize: bool = True,
+        normalize: str = "quantselect",


normalization_method: str | None = "directlfq"

removed normalize.

mschwoer · 2025-10-27T09:10:23Z

+            lfq_df = lfqutils.index_and_log_transform_input_df(intensity_df)
+            lfq_df = lfqutils.remove_allnan_rows_input_df(lfq_df)
+
+            if normalize == "directLFQ":


please put the strings directLFQ and quantselect into a class

class NormalizationMethods(metaclass=ConstantsClass): DIRECTLFQ: str = "directlfq" ...

and use it throughout the code?

done, moved class

class NormalizationMethods(metaclass=ConstantsClass): String constants for LFQ normalization methods. DIRECT_LFQ: str = "directLFQ" QUANT_SELECT: str = "QuantSelect"

to key.py

mschwoer · 2025-10-27T09:13:58Z

 psutil==7.0.0
 pyahocorasick==2.1.0
-pyarrow==19.0.1
+pyarrow==20.0.0


quantselect=0.1.0 is missing here,
also probably some others?

this file should anyway be autogenerated .. maybe we add the quantselect dependency (together with a bump of all requirement version) in a dedicated PR, @GeorgWa ?

mschwoer · 2025-10-27T09:15:10Z

+
+
+@pytest.fixture
+def create_psm_file():


(nit) as it's a fixture, should be called as what it is, e.g. test_psm_df

mschwoer · 2025-12-19T16:46:36Z

this can be closed now I guess @vuductung ? if so, please also delete the associated branch

vuductung added 14 commits October 22, 2025 15:27

Update README to enhance label free quantification description with m…

0a8a34b

…ultiple normalization methods (directLFQ, quantselect)

Add advanced quantselect normalization settings to default.yaml, incl…

8506aac

…uding model, training, optimizer, and prediction configurations.

Remove advanced quantselect normalization settings from default.yaml …

9167000

…to simplify configuration for users.

Add quantselect dependency to requirements.txt for enhanced normaliza…

098b7be

…tion support

Refactor label-free quantification process to incorporate normalizati…

65c1ecb

…on method directly in the intensity filtering step. Updated logging for fragment absence and adjusted input handling for lfq method to use the feature dictionary.

Refactor QuantBuilder to streamline imports and enhance code readabil…

28d65cb

…ity. Improved logging format for normalization processes and adjusted DataFrame handling in unit tests for better clarity and maintainability.

Update pyarrow version in requirements to 20.0.1 for compatiability w…

a0e3a06

…ith quantselect.

Downgrade pyarrow version in requirements to 20.0.0 to resolve compat…

dceda6c

…ibility issues.

Update normalization method in unit tests to use 'quantselect' instea…

a90d92f

…d of 'normalize_lfq' for consistency with recent changes in the quantification process.

Refactor accumulate_frag_df_from_folders method in QuantBuilder to re…

86b7370

…turn a dictionary instead of a tuple, improving clarity in handling fragment data. Update SearchPlanOutput to handle cases where no fragment data is found, enhancing logging for better debugging.

Refactor accumulate_frag_df_from_folders method in QuantBuilder to re…

f37057a

…turn a tuple instead of a dictionary, clarifying the handling of empty data cases. Update SearchPlanOutput to remove redundant checks for fragment data absence, improving code efficiency.

mschwoer reviewed Oct 27, 2025

View reviewed changes



		@pytest.fixture
		def create_psm_file():

Conversation

vuductung commented Oct 23, 2025

Uh oh!

mschwoer left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mschwoer commented Dec 19, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants