Overview The trifid.visualization.figures module provides plotting functions for visualizing TRIFID predictions, feature importance, and model performance. This is an optional module that requires visualization dependencies.
Install visualization dependencies with: pip install .[extra]
Installation # Install with visualization support pip install git+https://github.com/fpozoc/trifid.git pip install .[extra]
Required packages: altair, matplotlib, seaborn, shap, yellowbrick
Prediction Visualization explain_prediction() Generate SHAP force plot explanation for a specific transcript prediction.
DataFrame containing features and predictions
Trained TRIFID model object
List of feature names to use in explanation
Transcript ID to explain (e.g., “ENST00000356207”)
Returns: pd.DataFrame with SHAP valuesExample: from trifid.visualization.figures import explain_prediction # Load model and data model = load_model( 'path/to/model.pkl' ) df = pd.read_csv( 'predictions.tsv.gz' , sep = ' \t ' ) # Explain prediction for specific transcript explain_prediction( df = df, model = model, features = feature_list, transcript_id = 'ENST00000356207' )
This generates a SHAP force plot showing:
Base SHAP value (model’s expected value)
Feature contributions pushing prediction higher (orange)
Feature contributions pushing prediction lower (blue)
Final predicted score
plot_trifid_appris() Compare TRIFID scores against APPRIS annotations.
DataFrame with trifid_score and appris columns
Returns: Altair chart objectfrom trifid.visualization.figures import plot_trifid_appris df = pd.read_csv( 'predictions.tsv.gz' , sep = ' \t ' ) chart = plot_trifid_appris(df) chart.save( 'trifid_vs_appris.html' )
plot_appris_histogram() Create histogram of TRIFID scores grouped by APPRIS labels.
DataFrame with predictions and APPRIS annotations
Returns: Altair chart objectfrom trifid.visualization.figures import plot_appris_histogram, cat_appris_order # Prepare data with sorted APPRIS categories df = cat_appris_order(df) # Generate histogram chart = plot_appris_histogram(df) chart.save( 'appris_histogram.html' )
plot_transcript_types_histogram() Histogram of transcript type distribution.
DataFrame with transcript type flags
Returns: Altair chart objectfrom trifid.visualization.figures import plot_transcript_types_histogram, cat_transcript_type # Categorize transcript types df = cat_transcript_type(df) # Plot distribution chart = plot_transcript_types_histogram(df)
Feature Analysis plot_feature_importances() Visualize feature importance across multiple methods.
DataFrame with importance scores
Column name for x-axis (feature names)
Column name for y-axis (importance values)
Column for faceting (importance method)
Specific method to highlight
Number of top features to display
Returns: Altair chart objectfrom trifid.visualization.figures import plot_feature_importances # Create importance DataFrame importance_df = pd.DataFrame({ 'feature' : feature_names, 'importance' : shap_values, 'method' : 'SHAP' }) chart = plot_feature_importances( source = importance_df, xcol = 'feature' , ycol = 'importance' , facetcol = 'method' , method = 'SHAP' , ntop = 20 )
plot_pulse_comparison() Compare score distributions between two sets (e.g., TRIFID vs baseline).
DataFrame with scores and comparison labels
x_col
str
default: "trifid_score"
Column name for score values
x_axis_name
str
default: "TRIFID score"
Label for x-axis
Returns: Altair chart objectfrom trifid.visualization.figures import plot_pulse_comparison chart = plot_pulse_comparison( df = comparison_df, x_col = 'trifid_score' , x_axis_name = 'Functional Score' )
Model Evaluation plot_learning_curve() Generate learning curve to assess model performance vs training size.
scoring
str
default: "matthews_corrcoef"
Scoring metric
from trifid.visualization.figures import plot_learning_curve import matplotlib.pyplot as plt fig, ax = plt.subplots( figsize = ( 10 , 6 )) plot_learning_curve( model = rf_model, X = X_train, y = y_train, cv = 5 , scoring = 'matthews_corrcoef' , n_jobs =- 1 , ax = ax, title = 'TRIFID Learning Curve' ) plt.tight_layout() plt.savefig( 'learning_curve.png' , dpi = 300 )
plot_validation_curve() Plot validation curve for hyperparameter tuning.
Range of parameter values
scoring
str
default: "matthews_corrcoef"
Metric
from trifid.visualization.figures import plot_validation_curve fig, ax = plt.subplots( figsize = ( 10 , 6 )) plot_validation_curve( model = rf_model, X = X_train, y = y_train, param_name = 'n_estimators' , param_range = [ 50 , 100 , 200 , 400 , 600 ], cv = 5 , ax = ax, title = 'Validation Curve: n_estimators' )
plot_prcurve() Generate precision-recall curve with cross-validation.
from trifid.visualization.figures import plot_prcurve fig, ax = plt.subplots( figsize = ( 8 , 8 )) plot_prcurve( model = model, X = X_test, y = y_test, n_splits = '10' , seed = 42 , ax = ax, title = 'Precision-Recall Curve' )
Utility Functions cat_appris_order() Sort APPRIS annotations in canonical order.
DataFrame with appris column
Returns: DataFrame with appris_order column addedfrom trifid.visualization.figures import cat_appris_order df = cat_appris_order(df) # Now df has appris_order: 1-5 for PRINCIPAL, 6-7 for ALTERNATIVE, 8 for MINOR
cat_transcript_type() Categorize transcript type flags into readable labels.
DataFrame with flags column
Returns: DataFrame with flags_mod columnfrom trifid.visualization.figures import cat_transcript_type df = cat_transcript_type(df) # Adds flags_mod: "Protein coding", "Nonsense mediated decay", etc.
config_altair() Apply standard Altair chart configuration.
Returns: Configured Altair chartfrom trifid.visualization.figures import config_altair import altair as alt chart = alt.Chart(df).mark_bar().encode( x = 'feature' , y = 'value' ) chart = config_altair(chart, height = 400 , width = 600 )
create_categories() Create feature categories for comparative visualization.
Feature name to categorize
Returns: Categorized DataFramefrom trifid.visualization.figures import create_categories categorized = create_categories( df_features = features, df_predictions = predictions, feature = 'RNA2sj' , cats = [ 0 , 0.25 , 0.5 , 0.75 , 1.0 ] )
Complete Example import pandas as pd import pickle import matplotlib.pyplot as plt from trifid.visualization.figures import ( explain_prediction, plot_feature_importances, plot_appris_histogram, plot_learning_curve, cat_appris_order ) # Load data predictions = pd.read_csv( 'trifid_predictions.tsv.gz' , sep = ' \t ' ) with open ( 'trifid_model.pkl' , 'rb' ) as f: model = pickle.load(f) # 1. Explain a specific prediction explain_prediction( df = predictions, model = model, features = feature_list, transcript_id = 'ENST00000356207' ) # 2. Compare TRIFID vs APPRIS predictions = cat_appris_order(predictions) appris_chart = plot_appris_histogram(predictions) appris_chart.save( 'appris_comparison.html' ) # 3. Feature importance importance_chart = plot_feature_importances( source = importance_df, xcol = 'feature' , ycol = 'shap_value' , facetcol = 'method' , method = 'SHAP' , ntop = 15 ) importance_chart.save( 'feature_importance.html' ) # 4. Learning curve fig, ax = plt.subplots( figsize = ( 10 , 6 )) plot_learning_curve(model, X_train, y_train, ax = ax) plt.savefig( 'learning_curve.png' , dpi = 300 )
Source Reference Source: ~/workspace/source/trifid/visualization/figures.py
Related notebook: 02.figures.ipynb
See Also
Model Interpretation SHAP-based model interpretation functions
Interpreting Results Guide Step-by-step guide to result interpretation
