Addressing erroneous scale assumptions in microbe and gene set enrichment analysis.

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• Sequence count data, often compositional, lack system scale information.
• Differential Set Analysis (DSA) methods rely on normalization, making implicit scale assumptions.
• Errors in scale assumptions can drastically reduce the accuracy of DSA, with predictive values as low as 9%.

Purpose of the Study:

• To address scale limitations in Differential Set Analysis (DSA) of sequence count data.
• To develop methods for robust and reproducible scale-reliant inference.
• To guide researchers in evaluating the impact of scale limitations on their specific scientific goals.

Main Methods:

• Introduction of a novel sensitivity analysis framework applicable to both simulated and real data.
• Development of a statistical test that controls Type-I error rates despite scale assumption errors.
• Discussion of how scale limitations impact research goals and provision of evaluation tools.

Main Results:

• Commonly used DSA normalization methods make strong, implicit assumptions about unmeasured system scales.
• Small errors in scale assumptions can lead to significantly reduced positive predictive values.
• The proposed sensitivity analysis and statistical test offer robust inference capabilities.

Conclusions:

• Scale limitations in sequence count data analysis can lead to practical inferential errors.
• Rigorous and reproducible scale-reliant inference is achievable through careful application of new methods.
• This work aims to stimulate further research into scale limitations and their impact on biological data analysis.