Researchers from NIH have developed a 22-item framework, published in Nature, that has identified the minimal polygenic risk score-related information that scientists should include in their studies.
Polygenic risk score
Predisposition to common diseases and traits arises from a complex interaction between genetic and non-genetic risk factors. In the past decade, there have been huge efforts to identify disease-associated genetic variants. Genome-wide association studies, in particular, have yielded summary statistics that describe the magnitude and statistical significance of associations. The associations from these studies can be combined to quantify genetic predisposition to a heritable trait. This information can then be used to conduct disease risk stratification or predict prognostic outcomes and response to therapies. Typically, information across many variants is combined by means of a weighted sum of allele counts. This is also known as polygenic risk scores (PRSs).
In the last decade, the landscape of genetic prediction studies has transformed, with significant developments in how PRSs are constructed and evaluated. However, there is notable heterogeneity in the application and reporting of these scores. This limits the translation of PRSs into clinical care. In order to meaningfully evaluate these scores and determine which ones can be used in clinical care, universal standards across the research community need to be put in place.
The PRS Reporting Standards (PRS-RS)
The Clinical Genome Resource (ClinGen) Complex Disease Working Group and the Polygenic Score (PGS) Catalog have jointly presented the Polygenic Risk Score Reporting Standards (PRS-RS). The PRS-RS is an expanded and updated set of reporting standards for PRSs that specifically address current research environments with advanced methodological developments to inform clinically meaningful reporting on the development and validation of PRSs in the literature. Within these standards, there is an emphasis on reproducibility and transparency throughout the development process. These standards specify the minimal criteria that need to be described in a manuscript to accurately interpret a PRS and reproduce throughout the PRS development process. Below we briefly summarise these standards.
Reporting on background
- Study type: Researchers must specify whether the study aims to develop and/or validate a PRS.
- Risk model purpose and predicted outcome: Researchers must specify what the risk model is intended to predict and the purpose. This specifically includes intended use, predicted outcome and currently available models.
Reporting on study population and data
- Study design and recruitment: Researchers must describe the study design, eligibility criteria, recruitment period and setting, and follow-up.
- Participant demographics and clinical characteristics: Researchers must include the distribution of demographic information in each dataset. At minimum this should include age, sex and other relevant characteristics to describe the study population.
- Ancestry: Researchers must include the ancestral background distribution of each sample population used during PRS development and validation.
- Genetic data: Researchers must provide the method for acquiring genetic information in each sample. This includes information regarding genome build and technical assay details.
- Non-genetic variables: Researchers should define any non-genetic variables that were included in the risk model.
- Outcome of interest: Researchers should define the predicted outcome(s) of interest and report distribution. If the outcome is a clinical feature or endpoint within a specific disease, researchers should also provide the criteria used to define that disease membership.
- Missing data: Researchers must explicitly state how missing data were handled.
Reporting on risk model development and application
- PRS construction and estimation: Researchers should include how genetic data were included in the PRS, detailing the criteria used to determine inclusion in the model.
- Risk model type: Researchers should detail statistical methods used to estimate risk.
- Integrated risk model(s) description and fitting: Researchers should state the procedure used to develop the risk models that includes non-genetic and/or genetic variables other than the PRS.
Reporting on risk model evaluation
- PRS distribution: Researchers must include a general description of the distribution of the PRS.
- Model predictive ability: Researchers need to describe and report metrics of overall performance and estimates of risk used to evaluate the PRS.
- Risk model discrimination: Researchers need to describe and report metrics used to assess the discrimination of evaluated risk models and whether any non-genetic variables were included beyond a PRS in this analysis.
- Risk model calibration: Researchers should describe and report metrics used to assess the calibration of evaluated risk scores and models.
- Subgroup analyses: Researchers should provide information regarding subgroup size, demographics and clinical characteristics.
Reporting on limitations and clinical implications
- Risk model interpretation: Researchers must summarise the risk models in terms of what they predict, how well and in whom.
- Limitations: Researchers must outline limitations of the study with relevance to the results. They should also discuss the effects of these limitations on the interpretation of the risk model and any downstream replication efforts needed.
- Generalisability: Researchers should discuss the intended target groups or populations this score may be applied to. They should explicitly address any issues regarding generalisability beyond the included populations.
- Risk model intended uses: Researchers should mention whether there is an intended clinical use or utility to the risk model. If so, the authors should discuss the clinical readiness of the model and next steps.
Reporting on data transparency and availability
- Researchers should make sure information to calculate the PRS and the risk model(s) is freely available.
Without a consistent way of reporting PRSs, it is almost impossible to compare the utility of the scores. The PRS-RS enables the rapid development of PRSs that are potentially powerful clinical tools and also provides a framework for PRSs to transform multiple areas of genomic research. The team hope that if researchers follow these guidelines, evaluating published PRSs will be more straightforward, enabling experts to determine which ones are most appropriate for clinical settings.
Image credit: By Science Photo Library – canva.com