Accurate variant effect estimation in FACS-based deep mutational scanning data with Lilace.

Area of Science:

• Genomics
• Biophysics
• Computational Biology

Background:

• Deep mutational scanning (DMS) with fluorescence-activated cell sorting (FACS) is a powerful high-throughput technique.
• Linking genetic variants to molecular phenotypes using DMS-FACS data presents analytical challenges due to measurement variance and complex readouts.
• Existing statistical methods are insufficient for analyzing the multidimensional data generated by FACS-based DMS.

Purpose of the Study:

• To introduce Lilace, a novel Bayesian statistical model.
• To provide uncertainty quantification for variant effects estimated from FACS-based DMS experiments.
• To address the analytical challenges in analyzing DMS-FACS data.

Main Methods:

• Development of Lilace, a Bayesian statistical model.
• Validation using simulated datasets.
• Application to real-world DMS datasets (OCT1 and Kir2.1).

Main Results:

• Lilace effectively estimates variant effects with associated uncertainty.
• The model demonstrates robustness and improved performance on simulated data.
• Application to OCT1 and Kir2.1 datasets showed a reduced false discovery rate while maintaining sensitivity.

Conclusions:

• Lilace offers a robust statistical framework for analyzing FACS-based DMS experiments.
• The model enhances the accuracy of variant effect estimation and uncertainty quantification.
• Lilace represents a significant advancement in the analysis of high-throughput genetic variant data.