bio-temporal-genomics-trajectory-modeling

Version Compatibility

Reference examples tested with: R stats (base), numpy 1.26+, pandas 2.2+, scanpy 1.10+

Before using code patterns, verify installed versions match. If versions differ:

• Python: pip show <package> then help(module.function) to check signatures
• R: packageVersion('<pkg>') then ?function_name to verify parameters

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Temporal Trajectory Modeling

"Fit smooth curves to my gene expression time series" → Model continuous temporal trajectories using generalized additive models (GAMs) or spline regression, test for condition differences, and detect changepoints where dynamics shift abruptly.

• R: mgcv::gam() for GAM fitting with smooth terms
• Python: ruptures for changepoint detection in temporal profiles

Fits smooth non-linear curves to gene expression time series using generalized additive models (GAMs) and detects abrupt changes in temporal dynamics using changepoint algorithms.

Core Workflow

1. Prepare expression data with timepoint and condition metadata
2. Fit GAM or spline models per gene
3. Test for significant temporal trends and condition differences
4. Detect changepoints where trajectory dynamics shift
5. Predict and visualize fitted trajectories with confidence intervals

mgcv GAM (R)

Goal: Fit smooth non-linear curves to gene expression time series and test for significant temporal trends or condition differences.

Approach: Use generalized additive models with penalized smooth terms to capture non-linear dynamics, compare trajectories between conditions using interaction smooths, and extract predicted values with confidence intervals.

Basic GAM Fitting (R stats (base)+)

library(mgcv)

# gam() with s() smooth terms for non-linear temporal dynamics
# k=6: basis dimension (number of knots); controls smoothness
# k should be less than the number of unique timepoints
# Too low k = underfit; too high k = overfit; k=6 is a good default for 8-12 timepoints
fit <- gam(expression ~ s(time, k = 6), data = gene_df, method = 'REML')

summary(fit)
# edf (effective degrees of freedom): edf near 1 = linear; edf near k-1 = highly non-linear
# p-value for s(time): tests whether the smooth term is significantly non-zero