Published January 14, 2026 | Version v2

Genetically Instrumented CYP19A1 Whole Blood Expression and Autism Spectrum Disorder Risk: A Two-Sample Mendelian Randomization Study

  • 1. ROR icon Boston University

Description

This repository contains the analytic code, data outputs, and manuscript files for a two-sample Mendelian randomization study evaluating whether genetically predicted CYP19A1 (aromatase) whole blood expression is causally associated with autism spectrum disorder (ASD) risk.

Exposure: Three conditionally independent GTEx v8 whole blood cis-eQTL instruments (ENSG00000137869.14; GTEx Portal REST API v2). Outcome: iPSYCH-PGC ASD GWAS (Grove et al. 2019; ieu-a-1185; 18,382 cases, 27,969 controls).

Principal finding: The primary IVW estimate (OR = 1.131, 95% CI 1.002–1.277, p = 0.046) is not robust — it is absent in all six sensitivity analyses and driven by a sub-genome-wide significant instrument (rs75015950) with multi-gene eQTL effects at the chr15 locus. No genome-wide significant CYP19A1 cis-eQTLs were identified in any of 13 GTEx v8 brain tissues. This study establishes the feasibility constraints of current CYP19A1 expression instrumentation for ASD.

Files included: manuscript PDF, analytic Python scripts, mr_instruments.json, mr_results.json, phewas_summary.json, tissue_scan_results_complete.csv, requirements.txt.

Submitted to: BMC Medical Genomics (under review).
Reporting standard: STROBE-MR (Skrivankova et al. 2021, JAMA).

Notes

Our question?

Do currently available genetic instruments for CYP19A1 expression support a causal role in ASD risk?

Not the estradiol deficiency hypothesis itself — that is broader and older. Not whether aromatase matters in fetal brain — that is untestable with these data. The question actually answered is: given the genetic architecture of CYP19A1 expression as captured in GTEx v8, is there a detectable causal signal in the best available ASD GWAS? The answer is no.

Why is it relevant?

Three distinct gaps, in decreasing order of importance.

The first is empirical. Despite decades of interest in the estradiol deficiency hypothesis, no one had formally tested it in a genetic epidemiology framework. The Liu et al. 2025 CYP21A2 finding created an immediate scientific adjacency — if the androgen synthesis step upstream shows a causal signal, does the aromatization step downstream? This study is the first to ask that question with genetic instruments.

The second is methodological. The tissue scan across all 54 GTEx v8 tissues for CYP19A1 eQTLs was not previously published in any systematic form. The finding that zero GWS brain eQTLs exist for CYP19A1 in GTEx v8 — documented rigorously with full pagination and exact counts — is itself a data contribution. Any future researcher attempting this analysis will encounter the same architectural constraint. This study documents it explicitly so they do not have to rediscover it.

The third is interpretive. The study establishes a feasibility ceiling. It answers: what instrumentation does CYP19A1 MR for ASD require, and does it currently exist? The answer — brain-specific or fetal cortical eQTLs, not available in GTEx v8 — is a directional finding for the field, pointing toward BrainSeq Phase II, PsychENCODE, and dedicated fetal cortical eQTL datasets as the path forward.

 

The scientific contribution?

The contribution is a well-documented negative feasibility study that prevents wasted effort and redirects future work. 

More specifically, the contribution has four components.

Component

Content

Empirical finding

No robust causal signal for CYP19A1 whole blood expression on ASD risk across six sensitivity approaches including a pleiotropically bounded single-instrument analysis

Architectural finding

Zero GWS CYP19A1 cis-eQTLs across all 13 GTEx v8 brain tissues — the first systematic documentation of this constraint for this gene

Instrument characterization

rs75015950 identified and characterised as a dominant sub-GWS pleiotropic driver — a specific methodological finding about the available instrument set that would affect any future study using these variants

Feasibility map

Explicit delineation of what data would be required to actually test the estradiol deficiency hypothesis in a valid MR framework, with the specific resources named

 

Limitations

The contribution is bounded. This study cannot and does not test the biological hypothesis it is motivated by. The estradiol deficiency hypothesis concerns fetal brain aromatase activity. The instruments are from adult whole blood. The directional inconsistency of the nominal IVW result — OR > 1 when OR < 1 is predicted — is uninterpretable because the tissue mismatch means the instruments may not even be proxies for the relevant biological quantity.

Files

coloc_results_cyp19a1_asd.json

Files (1.6 MB)

Name Size
md5:a0e2fd34a470d107f047f3cc751d3c86
1.2 kB Preview Download
md5:cf8e6ed178dfb3e5f175e6671e8fbc9b
729.8 kB Download
md5:662a6ba30d0d151841e31a8ff831a77d
19.4 kB Download
md5:782bf4a1f0edd0475d45f7f61a64a27b
204.0 kB Preview Download
md5:69fd743b67d3d4a6a36f28cf7369c657
217.6 kB Preview Download
md5:487f74371d5405a28542955e190f21ce
377.3 kB Preview Download
md5:eeff04f25aa2aed54c4fd11983cf2c47
3.4 kB Preview Download

Additional details

Related works

Is supplement to
Software: https://github.com/julian-borges-md/frontier-research-lab (URL)
Is supplemented by
Journal article: 40307628 (PMID)
Reviews
Journal article: 10.1001/jama.2021.18236 (DOI)

Software