IsoFinder

IsoFinder predicts isochores in genomic sequences by segmenting DNA into long homogeneous genome regions (LHGRs) based on G+C content differences.

Key Features:

• Sliding-pointer segmentation: Employs a sliding pointer that traverses the sequence left to right and computes mean G+C values on either side of each position.
• t-statistic cut-point detection: Identifies potential cutting points where the difference between side mean G+C values is maximized using the t-statistic.
• Coarse-graining filter: Applies a coarse-graining technique to filter out short-scale heterogeneities below 3 kb.
• Statistical thresholding: Assesses significance of each cutting point against a predefined probability threshold and splits only when the threshold is exceeded.
• Recursive decomposition: Recursively analyzes resulting subsequences until the chromosome is decomposed into LHGRs with distinct mean G+C contents.
• Biological correlations: Produces isochores consistent with Bernardi's definition that correlate with gene density, SINE and LINE densities, recombination rates, and single nucleotide polymorphism variability.
• Isochore map generation: Generates isochore maps suitable for integration with genome resources such as the UCSC Genome Browser.

Scientific Applications:

• Genomic structure-function analysis: Relates LHGR composition to gene density, repeat element distributions (SINEs, LINEs), recombination rates, and SNP variability.
• Evolutionary genomics: Investigates isochore patterns and compositional domain evolution across genomes.
• Genome annotation and visualization: Provides compositional domain maps for integration with genome browsers and annotation efforts.

Methodology:

Traverse the sequence left to right with a sliding pointer, compute mean G+C on each side of each position, identify cut points maximizing the difference via the t-statistic, apply coarse-graining to remove heterogeneities below 3 kb, assess significance against a predefined probability threshold, and recursively segment subsequences into LHGRs.