Polygenic Prediction: Part 4 Bayesian methods for PGS prediction

Value of the Information & Strength of the Argument

The video provides a comprehensive and well-structured introduction to Bayesian methods for PGS prediction. It clearly explains the theoretical foundations, including the Bayes theorem and prior distributions, and connects them to practical algorithms like MCMC. The argumentation is solid, using examples and comparisons to illustrate the advantages of Bayesian approaches over BLUP. The instructor also addresses practical challenges, such as convergence and LD estimation, which adds depth and realism. The content is highly valuable for researchers or students in statistical genetics seeking a conceptual understanding of these methods.

The title accurately reflects the content, which focuses on Bayesian methods for polygenic score prediction.

8/10

The video is a technical lecture from an academic workshop, presenting established statistical methods (Bayesian alphabet, MCMC) with clear explanations and references to key literature. The content is accurate and well-structured, though it lacks formal citations within the video itself.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction to Bayesian methods for polygenic prediction and comparison with BLUP.
• 02:00Explanation of the infinitesimal model and its limitations.
• 04:00Introduction of BayesA and the use of t-distribution for SNP effects.
• 06:00Overview of Bayesian alphabet methods (BayesB, BayesC, BayesR) and their features.
• 08:00Explanation of Bayes theorem and its application in estimating parameters.
• 10:00Example of estimating average height using prior and likelihood.
• 12:00Derivation of posterior distribution and connection to BLUP.
• 14:00Introduction to MCMC algorithm for sampling from posterior distribution.
• 16:00Discussion of trace plots, burn-in, and posterior inclusion probability (PIP).
• 18:00Transition to summary-statistics-based methods and their advantages.
• 20:00Derivation of summary-level model from individual-level model.
• 22:00Discussion of convergence issues and practical recommendations.

• Landmark paper on BayesA — Mentioned as the earliest Bayesian model for genomic prediction.
• PRS-CS — Mentioned as a continuous shrinkage prior method.
• LDpred2 — Mentioned as a spike-and-slab or normal mixture model.
• SBayesR — Mentioned as a multi-component mixture model.

• Meuwissen et al. (2001) Prediction of total genetic value using genome-wide dense marker maps — Foundational paper introducing BayesA and BayesB.
• Vilhjálmsson et al. (2015) Modeling Linkage Disequilibrium Increases Accuracy of Polygenic Risk Scores — Introduces LDpred, a Bayesian method for PGS.
• Lloyd-Jones et al. (2019) Improved polygenic prediction by Bayesian multiple regression on summary statistics — Introduces SBayesR, a summary-statistics-based Bayesian method.