Genomic Data Smoothing

LOWESS for methylation profiles, ChIP-seq signals, and other genomic data.

Overview

Genomic data often contains noise from sequencing depth variation, PCR artifacts, or biological heterogeneity. LOWESS smoothing helps reveal underlying patterns.


Methylation Profile Smoothing

The Challenge

DNA methylation data (from bisulfite sequencing or arrays) shows position-dependent patterns that can be obscured by measurement noise.

Solution

A small fraction = 0.1 lets LOWESS follow fine-scale spatial structure without smearing the transitions between methylated and unmethylated regions. intervals = (confidence = 0.95,) produces uncertainty bands that naturally widen at positions with sparser CpG coverage, making low-confidence segments immediately apparent in the plot.

using FastLOWESS

positions = collect(0.0:1000.0:99000.0)
observed = 50.0 .+ sin.(positions ./ 1000.0) .* 20.0 .+ 5.0

# positions and observed are your methylation data
model = Lowess(;
    fraction=0.1,
    iterations=3,
    intervals=(confidence=0.95,),
)
result = fit(model, positions, observed)

# Smoothed profile in result.y
# CI bounds in result.confidence_lower/upper
println("95% CI: [", result.confidence_lower[1], ", ", result.confidence_upper[1], "]")
95% CI: [53.27919505033735, 67.15544006032717]

ChIP-seq Signal Smoothing

Application

ChIP-seq experiments produce sparse, noisy coverage data. LOWESS can help identify binding regions.

fraction = 0.05 provides high spatial resolution — important for resolving narrow binding peaks that would otherwise be smeared into the background. The larger iterations = 5 is deliberate: Poisson-distributed read counts produce tall, isolated spikes, and extra robustness iterations progressively down-weight them so the estimated background level is not inflated by a handful of extreme counts.

using FastLOWESS

positions = collect(0.0:1000.0:99000.0)
observed = 50.0 .+ sin.(positions ./ 1000.0) .* 20.0 .+ 5.0

# positions and observed are your ChIP-seq data
model = Lowess(; fraction=0.05, iterations=5)
result = fit(model, positions, observed)

# Find peaks above threshold
peak_count = count(y -> y > 65.0, result.y)
println("y[0]: ", result.y[1])
println("Peak count: ", peak_count)
y[0]: 59.937594476093395
Peak count: 26

Large Genome Coverage (Streaming)

For whole-genome data that doesn't fit in memory:

using FastLOWESS

positions = collect(0.0:10.0:10000.0)
coverage = 50.0 .+ sin.(positions ./ 100.0) .* 20.0 .+ 5.0

# coverage and positions are chromosome-scale vectors
model = StreamingLowess(;
    fraction=0.05,
    iterations=3,
    chunk_size=50,
    overlap=10,
    merge_strategy="weighted_average"
)
process_chunk(model, positions, coverage)
result = finalize(model)
println("y[0]: ", result.y[1])
y[0]: 41.28390556612269

Best Practices for Genomic Data

ConsiderationRecommendation
Fraction0.05–0.15 (preserve local features)
Iterations3–5 (handle sequencing outliers)
Large dataUse streaming mode
Sparse regionsUse boundary_policy="extend"
Multiple chromosomesProcess separately or ensure sorted

See Also