Accurate and fast small p-value estimation for permutation tests in high-throughput genomic data analysis with the...

Area of Science:

• Genomics
• Computational Biology
• Statistical Genetics

Background:

• Permutation tests are crucial for hypothesis testing when distributions are complex.
• Estimating very small p-values in genomic studies requires extensive computation.
• Existing methods face computational challenges with large genomic datasets.

Purpose of the Study:

• To develop accurate and efficient algorithms for estimating small p-values in permutation tests.
• To address the computational intensity of permutation tests in genomic data analysis.
• To provide improved solutions for hypothesis testing in genomics.

Main Methods:

• Developed novel algorithms for paired and independent two-group genomic data.
• Utilized Bernoulli and conditional Bernoulli distributions to parameterize sample spaces.
• Employed the cross-entropy method for efficient estimation.
• Leveraged novel frameworks for parameterizing permutation sample spaces.

Main Results:

• Achieved orders of magnitude computational efficiency gains in estimating small p-values.
• Demonstrated performance on simulated and real-world gene expression datasets (microarray, RNA-Seq).
• Outperformed existing methods like crude permutations and SAMC in efficiency.

Conclusions:

• Proposed algorithms offer significant computational advantages for genomic permutation tests.
• These methods enhance the efficiency of existing permutation test procedures.
• The framework facilitates the development of new permutation-based genomic analysis tools.