LRMotifs

LRMotifs applies logistic regression and hypothesis testing to discover DNA sequence motifs that indicate transcription factor binding sites and to predict gene expression and gene-cluster membership across cellular conditions.

Key Features:

• Logistic Regression-Based Approach: Employs logistic regression to model the probability of binary outcomes from DNA sequence features, such as membership in gene clusters or transcription factor (TF) binding.
• Rigorous Hypothesis Testing: Uses DNA background sequence models to represent the null hypothesis and test the statistical significance of discovered motifs.
• Unbiased Validation: Assesses motif predictive performance on held-out data to validate motifs and reduce overfitting.
• De Novo Motif Discovery: Searches sequences without prior knowledge of binding sites or nucleotide patterns.
• Statistical Intractability and Prior Knowledge: Recognizes statistical challenges from fixed sizes of co-regulated gene clusters and emphasizes incorporating prior knowledge when necessary.

Scientific Applications:

• Reverse-engineering cis-regulatory logic: Infers cis-regulatory motifs underlying control of gene expression.
• Transcription factor binding site identification: Identifies sequence motifs that serve as TF binding sites.
• Predictive modeling of gene expression and cluster membership: Predicts expression patterns and membership in co-regulated gene clusters across cellular conditions.

Methodology:

De novo motif discovery; logistic regression modeling of motif associations with binary outcomes; statistical significance testing using real DNA background sequence models; and evaluation of predictive performance on held-out data.