DIMA

DIMA selects the optimal imputation algorithm for proteomics datasets using data-driven performance evaluations to improve handling of missing values in quantitative proteomics analyses.

Key Features:

• Performance Assessment: DIMA evaluates the performance of multiple imputation algorithms across diverse quantitative proteomics datasets to identify high-performing methods tailored to dataset properties.
• Data-Driven Selection: Using empirical evaluations on 142 quantitative proteomics datasets from the PRoteomics IDEntifications (PRIDE) database and simulated datasets with 5-50% missing values, DIMA determines the most suitable imputation algorithm based on performance metrics.
• Robustness Across MV Types: DIMA handles missing not at random and missing completely at random scenarios to accommodate different missing-value mechanisms.
• Accuracy and Reliability: DIMA consistently identifies an algorithm among the top three performers and achieves a ΔRMSE ≤ 10% in 80% of evaluated cases.

Scientific Applications:

• Proteomics Data Analysis: Optimizing imputation strategies to mitigate the impact of missing values on quantitative proteomics results and downstream analyses.
• Broad Applicability: Applicable to datasets beyond proteomics by adapting to varying proportions and types of missing values.

Methodology:

DIMA evaluates multiple imputation algorithms on real-world (PRIDE) and simulated datasets and recommends the algorithm with the best empirical performance for the specific dataset characteristics.