Holden

Holden detects fabrication and falsification in large-scale molecular omics datasets using machine learning analyses to assess data authenticity.

Key Features:

• Raw Data Analysis: Accepts raw omics datasets as input and applies machine learning models to identify fabricated results.
• Digit Frequency Analysis: Uses digit frequency within datasets as input features to identify unnatural patterns associated with fabrication, with reported detection rates of 98%–100%.
• Machine Learning Detection Performance: Machine learning models applied to raw data report 84%–95% success rates in identifying fabricated results.
• Scope: Targets large-scale molecular omics datasets for integrity assessment and validation of data authenticity.

Scientific Applications:

• Fabrication and falsification detection: Identification of fabricated or falsified results within molecular omics datasets.
• Data authenticity validation: Validation of omics data integrity for downstream analyses and reproducibility assessments.

Methodology:

Inputs are raw omics datasets; machine learning models are applied to raw data; digit frequency within datasets is used as input features to machine learning models; reported detection rates are 84%–95% for raw-data models and 98%–100% for digit-frequency-based models.