8 AutoFeatureSelection.py

Run the complete analysis pipeline from data loading to training and result extraction in TransProPy.UtilsFunction3. This function streamlines the process of data handling, feature selection, and model evaluation, providing a comprehensive approach for machine learning pipelines.

8.1 pseudo-code

8.2 Parameters

data_file: str:

Path to the feature data file.

label_file: str:

Path to the label data file.

label_col: str:

Name of the label column.

threshold: float:

Threshold for data preprocessing.

show_plot: bool:

Whether to display plot.

show_progress: bool:

Whether to show a progress bar.

n_iter: int:

Number of iterations for RandomizedSearchCV.

n_cv: int:

Number of folds for cross-validation.

n_jobs: int:

Number of parallel jobs for RandomizedSearchCV.

save_path: str:

Path to save results.

sleep_interval: int:

Interval time in seconds for progress bar update.

use_tkagg: bool:

Whether to use ‘TkAgg’ backend for matplotlib. Generally, choose True when using in PyCharm IDE, and choose False when rendering file.qmd to an HTML file.

8.3 Please note

Note

If n_jobs > 1 leads to the use of multiprocessing in the function, communication and state sharing between processes become complex. In a multiprocessing environment, each process has its own memory space, and updating the status of the progress bar needs to be done across processes, which can lead to inconsistent or delayed updates. Therefore, if n_jobs > 1 or = -1, please set show_progress to False. If you have sufficient computing resources, such as on a cloud server, pay attention to whether the number of computing nodes, cores per node, and the number of CPUs allocated per task match the value of n_jobs.

8.4 Description

The auto_feature_selection function automates the entire process of feature selection, model training, and result extraction. It includes:

Loading and preprocessing data using load_and_preprocess_data.

Setting up feature selection with setup_feature_selection.

Defining and training a model with train_model using RandomizedSearchCV.

Optionally displaying a progress bar using tqdm and threading for real-time progress updates.

Extracting and saving results with extract_and_save_results.

Tip

Single-threaded computation: show_progress=True, n_jobs=1 Multithreaded computation: show_progress=False, n_jobs=n

8.5 Usage (Multithreaded computation: show_progress=False, n_jobs=-1)

8.5.1 four_methods_degs_intersection

from TransProPy.AutoFeatureSelection import auto_feature_selection
auto_feature_selection(
    data_file='../test_TransProPy/data/four_methods_degs_intersection.csv',
    label_file='../test_TransProPy/data/class.csv',
    label_col='class',
    threshold=0.95,
    show_plot=True,
    show_progress=False,
    n_iter=8,
    n_cv=3,
    n_jobs=-1,
    save_path='../test_TransProPy/result/four_methods_degs_intersection_',
    sleep_interval=100,
    use_tkagg=False
)

Fig1

Fig2

#=============================#
# Features selected by RFECV: #
#=============================#
Index(['A2M', 'A2ML1', 'ABCA17P', 'ABCA9', 'ABCB4', 'ABCD1', 'AC016757.3',
       'AC138035.2', 'AC159540.1', 'ACAD11',
       ...
       'ZBTB16', 'ZBTB32', 'ZBTB9', 'ZNF114', 'ZNF280B', 'ZNF337', 'ZNF410',
       'ZNF564', 'ZNF697', 'ZNRF3'],
      dtype='object', length=561)

#===================================#
# Features selected by SelectKBest: #
#===================================#
Index(['ABCD1', 'AC125232.1', 'AC138035.2', 'AC159540.1', 'ACAD11', 'ACOT7',
       'AGAP5', 'AKAP2', 'ALX1', 'ANKHD1',
       ...
       'TRIB2', 'TRIM34', 'TTC4', 'U2AF1', 'UBBP4', 'UBE2V1',
       'XXbac-B461K10.4', 'ZBTB9', 'ZNF410', 'ZNRF3'],
      dtype='object', length=184)

#========================================#
# Total number of selected features: 581 #
#========================================#

#==========================================#
# Feature Importances from EnsembleForRFE: #
#==========================================#
                    Feature  Importance
0             SLX1A-SULT1A3    0.455213
1                    EIF3CL    0.095509
2                     EEF1G    0.061155
3             RP11-231C14.4    0.060150
4                 RPL13AP25    0.001863
5                     MEX3A    0.001836
6                     SRXN1    0.001824
7                  EEF1A1P9    0.001793
8                PPP1R14BP3    0.001741
9             RP11-513I15.6    0.001741
10                HSD17B1P1    0.001626
11                 HNRNPKP1    0.001606
12                    TUBA8    0.001606
13                RPL13AP20    0.001594
14                 HNRNPCP2    0.001592
15                    PYURF    0.001572
16             RP11-676M6.1    0.001569
17                EEF1A1P12    0.001568
18           C15orf38-AP3S2    0.001516
19                     DLX1    0.001484
20            RP11-543P15.1    0.001480
21             RP3-475N16.1    0.001441
22                    WDR88    0.001439
23                    TIMP3    0.001431
24                   BOLA2B    0.001430
25             RP4-756H11.3    0.001423
26            RP11-977G19.5    0.001423
27             CTC-451P13.1    0.001399
28                    AKAP2    0.001398
29            CTD-2303H24.2    0.001397
..                      ...         ...
531                 IL12RB1    0.000578
532                 RASGRF1    0.000577
533           RP5-1039K5.19    0.000577
534  DTX2P1-UPK3BP1-PMS2P11    0.000574
535                   FCRLA    0.000569
536                    C1QC    0.000569
537                  ARL2BP    0.000565
538            RP11-434H6.6    0.000565
539                   KRT14    0.000564
540                   MCHR1    0.000564
541                  DBNDD1    0.000563
542                  ADAM23    0.000555
543                     ADM    0.000554
544                    CD74    0.000553
545                    LAG3    0.000553
546                 SLC28A3    0.000552
547                     KMO    0.000551
548                   KLK10    0.000551
549                   ITGAL    0.000550
550                   NEAT1    0.000549
551                   RSAD2    0.000537
552                  PKD1P5    0.000537
553                   SEPT3    0.000527
554                     LYZ    0.000525
555            RP11-20D14.6    0.000523
556                   GSDMC    0.000518
557                  FAM46C    0.000511
558                 SLCO5A1    0.000504
559                   GSDMB    0.000435
560                     FLG    0.000432

[561 rows x 2 columns]

#==========================#
# Scores from SelectKBest: #
#==========================#
            Feature     Score
0             EEF1G  0.657676
1     SLX1A-SULT1A3  0.657676
2            EIF3CL  0.656309
3     RP11-231C14.4  0.655948
4             U2AF1  0.654053
5              TEN1  0.653639
6           FAM156A  0.653524
7     CTD-2231E14.8  0.652795
8            NPIPB5  0.652436
9           GTF2IP1  0.651938
10   TMEM256-PLSCR3  0.651155
11           EIF4A1  0.650506
12          ARL6IP4  0.650506
13            ATRIP  0.649792
14            SPSB3  0.648970
15       PPP1R14BP3  0.648784
16   C15orf38-AP3S2  0.648361
17            ZBTB9  0.647219
18            AKAP2  0.646561
19           NPIPA1  0.646222
20            SNX15  0.646055
21        RPL13AP25  0.645636
22            AGAP5  0.645010
23            CBWD3  0.644778
24            EIF3C  0.644602
25           PIK3R2  0.644362
26           NPIPB3  0.643772
27            MATR3  0.643762
28           ZNF410  0.643468
29            TOP3B  0.643425
..              ...       ...
154           ABCD1  0.581264
155          DBNDD1  0.579869
156      AC138035.2  0.579813
157       HSD17B1P1  0.579566
158           CHST6  0.579342
159         RABGEF1  0.579211
160            PLP1  0.578937
161          TRIM34  0.578699
162    RP11-42I10.1  0.578196
163            ALX1  0.577680
164          PIK3CD  0.577326
165    RP4-756H11.3  0.576957
166           PLOD3  0.576935
167          FAM78A  0.576135
168            CHKB  0.575878
169             SDS  0.575652
170            CCL5  0.575628
171         CYP27A1  0.574835
172          RNF157  0.574751
173    RP11-419C5.2  0.574712
174        C19orf38  0.573959
175           NHEJ1  0.573535
176     RP3-337H4.8  0.573386
177          IGSF11  0.573195
178       RPL13AP20  0.573170
179   RP11-549B18.1  0.571859
180           HLA-G  0.571613
181     CH507-9B2.3  0.571400
182           ACOT7  0.570546
183           S100B  0.570479

[184 rows x 2 columns]

8.5.2 four_methods_degs_union

from TransProPy.AutoFeatureSelection import auto_feature_selection
auto_feature_selection(
    data_file='../test_TransProPy/data/four_methods_degs_union.csv',
    label_file='../test_TransProPy/data/class.csv',
    label_col='class',
    threshold=0.95,
    show_plot=True,
    show_progress=False,
    n_iter=8,
    n_cv=3,
    n_jobs=-1,
    save_path='../test_TransProPy/result/four_methods_degs_union_',
    sleep_interval=100,
    use_tkagg=False
)