Comprehensive Polygenic Score Profiling Reveals Autism Spectrum Disorder Subgroups with Different Genetic Predisposition Related to High-Density Lipoprotein Cholesterol, Urea, and Body Mass Index

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study profiles 2,602 polygenic scores in 75 autistic individuals, identifying three genetically distinct subgroups linked to lipid metabolism, renal function, and BMI. Findings suggest PGS-based stratification may clarify ASD heterogeneity, though limited sample size and ancestry constraints weaken generalizability. Here are some comments that may help the authors improve this manuscript.

 

Major comments

1. Provide formal statistical justification for selecting three clusters, including quantitative cluster validity indices (e.g., silhouette width, gap statistic, Dunn index).
2. Conduct cluster stability analyses such as bootstrapped Jaccard similarity to demonstrate that clustering is reproducible and not an artifact of high-dimensional PGS space.
3. Repeat clustering using alternative algorithms (e.g., k-means, Gaussian mixture models, spectral clustering) to confirm robustness of subgroup structure.
4. Re-run clustering on PCA-reduced data (e.g., top 20–50 PCs) to determine whether the clusters persist after dimensionality reduction.
5. Apply global false-discovery-rate correction across all PGS comparisons rather than correcting within each subgroup only.
6. Report effect sizes and confidence intervals for all subgroup PGS differences to enhance interpretability and reduce reliance on p-values alone.
7. Address the limited transferability of predominantly European-derived PGSs to a Japanese cohort and quantify expected variance explained in East Asian populations.
8. Conduct sensitivity analyses using PGSs derived from East Asian GWAS to evaluate whether cluster structure is ancestry-robust.
9. Account for correlation between PGS definitions within traits by pruning highly correlated PGSs (e.g., r > 0.8) and repeating enrichment analyses.
10. Include a correlation heatmap to document redundancy among PGSs contributing to trait enrichments.
11. Provide further details about how the interpretation of metabolic or renal predispositions was made, as no phenotypic measurements (e.g., HDL-C, urea, BMI) were available to validate these inferences.
12. Revise claims suggesting clinical utility, drug-development implications, or biomarker readiness to reflect the exploratory nature of this work.
13. Provide internal validation (e.g., cross-validation, leave-one-out cluster stability testing) to support reproducibility in the absence of an external dataset.
14. Clearly distinguish between pre-specified hypotheses and findings that emerged post hoc during clustering and enrichment analysis.
15. Expand the limitations section to address high-dimensional clustering instability, PGS ancestry mismatch, sample size constraints, lack of controls, and self-reported ASD diagnosis.

 

Minor Comments

1. Clarify how missing SNPs within PGS definitions were handled during scoring.
2. Provide a flow diagram showing all PGS definitions downloaded, excluded for licensing, excluded for QC, and retained for analysis.
3. Justify the criterion excluding PGSs with >30% identical values across samples.
4. Confirm that all PGSs were standardized prior to PCA and clustering.
5. Include variance explained by the top principal components in PCA visualizations.
6. Improve readability of heatmaps and scatter plots by simplifying labels and adopting color-blind–safe color schemes.
7. Supply a supplemental table listing all distinctive PGSs with effect sizes, confidence intervals, and original GWAS ancestries.
8. Provide clearer descriptions of demographic variables and any available clinical characteristics for each subgroup.
9. Define all acronyms at first use.
10. Improve figure resolution and ensure consistent formatting across all panels and supplemental materials.

Comments on the Quality of English Language

The manuscript is generally understandable, but the quality of English can be improved to enhance clarity and readability. Several sentences are overly long or grammatically awkward, and some technical descriptions would benefit from more precise wording. A careful language edit—focusing on conciseness, transitions, and scientific clarity—would strengthen the presentation and help readers better follow the methodology and interpretations.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript investigates the utility of polygenic scores (PGS) as biomarkers for stratifying individuals with autism spectrum disorder (ASD). Based on the premise that ASD is a clinically and genetically heterogeneous disorder, the authors propose that employing a comprehensive panel of PGSs can identify genetic subgroups within the autism spectrum, potentially advancing precision medicine and the development of tailored interventions.

It is recommended to inform the need to replicate the findings in independent cohorts and the importance of conducting longitudinal studies to evaluate the stability of the subgroups. Additionally, discuss potential biases and sources of error, such as selection bias, inaccuracies in genetic imputation, and confounding factors due to uncontrolled comorbidities.

Exercise caution in clinical interpretation, emphasizing that polygenic scores (PGSs), although promising, do not yet possess established diagnostic or prognostic utility in clinical practice for ASD.

Statistics

Implement cluster stability analyses, such as bootstrapping and cross-validation, to demonstrate the robustness of the stratification.

Please present some statistics to assess the quality and optimal number of clusters: elbow curve, silhouette score, gap statistic method, or Davis-Bouldin score.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you very much for your substantial and thoughtful revisions, which clearly reflect a great deal of effort and engagement with the review.​

One key point that still feels insufficiently addressed is the lack of formal cluster validity and stability assessment. In addition to the k-means sensitivity check and PCA visualizations you now provide, it would greatly strengthen the manuscript if you could implement a quantitative stability analysis (for example, a resampling or bootstrap-based assessment using Jaccard similarity or related indices) and, where feasible, report standard cluster validity metrics for the chosen solution. This would help move the three-cluster structure from a primarily visually supported finding to one with more formal empirical backing.

Author Response

We sincerely thank the reviewer for the encouraging assessment of our revised manuscript. We appreciate the reviewer’s valuable suggestion regarding the implementation of formal cluster validity and stability assessments. As we noted previously, our methodological framework relies on manual visual inspection for cluster determination, which does not support automated sample assignment and therefore limits the use of quantitative stability and validity metrics such as bootstrapped Jaccard similarity.

To address this limitation, we have already included an explanation in the Discussion section of the manuscript, outlining the rationale for our approach and explicitly acknowledging the limitation this imposes on formal validity assessment. Given this, we believe no further revision is necessary at this time. Therefore, we have retained our prior responses and discussion as they comprehensively cover the reviewer’s concerns within the context of our methodological choices.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript is well-written and well-supported. The authors addressed the feedback and suggestions. I have no further comments. Well done.

Author Response

Thank you very much for your positive and encouraging feedback.
We sincerely appreciate your careful review of our manuscript and are grateful that the revisions addressed your concerns.
Your comments were very helpful in improving the quality of our work.