LolliPop

LolliPop estimates relative abundances of genomic variants from time-series sequencing data of sewage to enable variant deconvolution for wastewater-based epidemiology.

Key Features:

• Kernel-Based Deconvolution: LolliPop applies a kernel-based deconvolution with temporal regularization implemented via a fused ridge penalty, which is equivalent to kernel smoothing for abundance estimation.
• Temporal Regularization: It incorporates temporal regularization to exploit the time-series structure of wastewater samples and improve tracking of variant dynamics.
• Handling Missing Data: The method is tailored to handle noisy and incomplete wastewater sequencing datasets with prevalent missing values.
• Confidence Interval Estimation: LolliPop generates confidence intervals using bootstrap resampling and provides analytical standard errors that approximate bootstrap intervals with reduced computational cost.
• Validation on Simulated and Real Datasets: The approach has been demonstrated on simulated data and on Swiss national variant monitoring datasets to provide early detection and unbiased estimates of viral loads and variants.

Scientific Applications:

• Wastewater-Based Epidemiology: Estimating circulating viral variant abundances from sewage sequencing to inform population-level surveillance and epidemiological models.
• SARS-CoV-2 Variant Surveillance: Early detection and unbiased quantification of SARS-CoV-2 variants as a complementary signal to clinical surveillance, demonstrated on national monitoring data from Switzerland.

Methodology:

Kernel-based deconvolution with temporal regularization via a fused ridge penalty (equivalent to kernel smoothing), bootstrap resampling for confidence intervals, and analytical standard-error estimation applied to time-series sequencing data from sewage.