Skip to contents

Build a sex prediction model from a NIPTeR control group

nipter_sex_model.Rd

Fits a two-component Gaussian mixture model (GMM) on sex chromosome fractions derived from a NIPTeRControlGroup. The model distinguishes male and female samples based on Y-chromosome read fraction (univariate) or X+Y chromosome fractions (bivariate).

Usage

nipter_sex_model(control_group, method = c("y_fraction", "xy_fraction"))

Arguments

control_group

A NIPTeRControlGroup object. Ideally chi-squared corrected via nipter_chi_correct.

method

Character; the feature space for the GMM:

"y_fraction"

Univariate: Y-chromosome read fraction relative to total autosomal reads.

"xy_fraction"

Bivariate: X and Y chromosome fractions.

Value

An object of class "NIPTeRSexModel" with elements:

model

The mclust::Mclust fitted object.

method

Character; the method used.

male_cluster

Integer (1 or 2); which cluster is male (higher Y fraction).

classifications

Named character vector of "male"/ "female" labels for each control sample.

fractions

Matrix of the input features used for fitting (samples as rows). Columns are "y_fraction" or c("x_fraction", "y_fraction").

Details

This mirrors the approach used in clinical NIPT pipelines (see Details), ported to R using mclust::Mclust() in place of Python's sklearn.GaussianMixture.

The algorithm:

  1. Compute sex chromosome read fractions for every sample in the control group. Y fraction = sum(Y bins) / sum(autosomal bins); X fraction = sum(X bins) / sum(autosomal bins).

  2. Fit a two-component Gaussian mixture with equal mixing proportions (mclust::Mclust(data, G = 2, control = mclust::emControl(equalPro = TRUE))).

  3. Identify the male cluster as the component with the higher median Y fraction.

This follows the clinical NIPT pipeline pattern of building sex models from control cohorts. The user's pipeline builds three models (Y-unique ratio from samtools, XY fractions, Y fraction) and takes a majority vote; nipter_predict_sex() implements the consensus when given multiple models.

See also

nipter_predict_sex(), nipter_chi_correct()

Examples

if (FALSE) { # \dontrun{
# Build sex model from chi-corrected control group
sex_model <- nipter_sex_model(chi_cg, method = "y_fraction")

# Bivariate X+Y model
sex_model_xy <- nipter_sex_model(chi_cg, method = "xy_fraction")
} # }