Skip to content

Latest commit

 

History

History
95 lines (65 loc) · 11 KB

on_boarding.md

File metadata and controls

95 lines (65 loc) · 11 KB
graph LR
    EFAAR_Data_Transformation["EFAAR Data Transformation"]
    Benchmarking_Core_Utilities["Benchmarking Core & Utilities"]
    Perturbation_Signal_Benchmarking["Perturbation Signal Benchmarking"]
    Relationship_Cluster_Benchmarking["Relationship & Cluster Benchmarking"]
    Compound_Gene_Interaction_Benchmarking["Compound-Gene Interaction Benchmarking"]
    EFAAR_Data_Transformation -- "Provides processed data for" --> Perturbation_Signal_Benchmarking
    EFAAR_Data_Transformation -- "Provides processed data for" --> Relationship_Cluster_Benchmarking
    EFAAR_Data_Transformation -- "Provides processed data for" --> Compound_Gene_Interaction_Benchmarking
    Benchmarking_Core_Utilities -- "Provides configuration and utilities for" --> Perturbation_Signal_Benchmarking
    Benchmarking_Core_Utilities -- "Provides configuration and utilities for" --> Relationship_Cluster_Benchmarking
    Benchmarking_Core_Utilities -- "Provides configuration and utilities for" --> Compound_Gene_Interaction_Benchmarking
    click EFAAR_Data_Transformation href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//EFAAR_benchmarking/EFAAR Data Transformation.md" "Details"
    click Benchmarking_Core_Utilities href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//EFAAR_benchmarking/Benchmarking Core & Utilities.md" "Details"
    click Perturbation_Signal_Benchmarking href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//EFAAR_benchmarking/Perturbation Signal Benchmarking.md" "Details"
    click Relationship_Cluster_Benchmarking href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//EFAAR_benchmarking/Relationship & Cluster Benchmarking.md" "Details"
    click Compound_Gene_Interaction_Benchmarking href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//EFAAR_benchmarking/Compound-Gene Interaction Benchmarking.md" "Details"
Loading

CodeBoardingDemoContact

Component Details

This architecture describes the EFAAR_benchmarking project, which focuses on evaluating the performance of the EFAAR method for analyzing biological perturbations. The main flow involves transforming raw biological data using various normalization and dimensionality reduction techniques, followed by comprehensive benchmarking against different criteria such as perturbation signal consistency, known biological relationships, cluster enrichment, and compound-gene interactions. A core utility component supports all benchmarking activities by providing configuration, truth data loading, and common computational functions.

EFAAR Data Transformation

Handles core data processing steps of the EFAAR method, including dimensionality reduction using PCA and various normalization techniques (batch, control-based scaling/centering, TVN) to prepare raw biological data for downstream analysis.

Related Classes/Methods:

Benchmarking Core & Utilities

Provides the foundational framework for benchmarking, including configuration settings, loading ground truth data, and general utilities for computing similarities and processing predictions across various benchmarks.

Related Classes/Methods:

Perturbation Signal Benchmarking

Evaluates the consistency and magnitude of perturbation signals within the transformed data, calculating metrics like average cosine similarity and energy distance, often against a null distribution.

Related Classes/Methods:

Relationship & Cluster Benchmarking

Assesses the recall of known biological relationships (e.g., compound-compound, gene-gene interactions) and evaluates performance based on known biological clusters of perturbations, including enrichment analysis.

Related Classes/Methods:

Compound-Gene Interaction Benchmarking

Manages the end-to-end process of predicting and evaluating compound-gene interactions, including processing predictions, sampling for items, computing average precision and AUC-ROC, and generating baseline predictions for comparison.

Related Classes/Methods:

FAQ