Extreme phenotypes and two-stage QTL annotation in hemoglobinopathy genetics

Question: How have multi-omic quantitative trait locus (QTL) integration approaches—such as SNP-protein and SNP-metabolite associations combined with pathway enrichment—been used to dissect genetic risk loci for severe recurrent phenotypes in hemoglobinopathies or other Mendelian diseases?

Answer

Across the retrieved literature, multi-omic QTL integration in Mendelian disease is used less as a locus-resolving instrument and more as a two-stage annotation-and-triage pipeline: first, an extreme or recurrent phenotype is defined to manufacture genetic signal that the ordinary phenotype definition cannot deliver; second, SNP-protein and SNP-metabolite associations computed against that phenotype are collapsed into gene sets for pathway enrichment, and pQTLs are repurposed as Mendelian randomization instruments to nominate causal proteins and drug targets. The single hemoglobinopathy study that executes the full SNP-protein + SNP-metabolite + pathway-enrichment sequence does so on acute chest syndrome recurrence, where the occurrence phenotype yielded no genome-wide significant SNP but the recurrence phenotype yielded one, and where the molecular QTL evidence sits at p<1e-3 and is used to describe rather than to prove mechanism [2]. The causal claims that survive scrutiny in these papers come from a different step—pQTL-instrumented Mendelian randomization and colocalization, as in the prioritization of five proteins predicted to raise fetal hemoglobin in sickle cell disease [1]. Taken together, the retrieved work supports multi-omic QTL integration as a hypothesis-generating and target-triage layer whose value is currently bounded by cohort size, by the near-total concordance of disease-context pQTL effect sizes with population pQTLs, and by unresolved questions about what colocalization evidence actually licenses.

Extreme recurrence phenotypes buy the genetic signal that occurrence phenotypes do not

The clearest demonstration is in sickle cell disease acute chest syndrome, where a meta-analysis of occurrence identified no SNP at genome-wide significance and 40 independent loci at p<1e-5, whereas a recurrence analysis contrasting no prior episodes against more than five episodes identified one genome-wide significant locus plus 38 independent loci at p<1e-5 [2]. Survival analysis in that study suggested lower survival probability among young recurrence cases compared with controls or occurrence cases, which is the justification offered for treating recurrence as the more informative phenotype for risk prediction and therapeutic targeting [2]. The same logic of sharpening the phenotype appears outside hemoglobinopathies, where restricting analyses to a defined extreme stratum (incomplete-penetrance families, calibrated proxy phenotypes) raised diagnostic or discovery yield relative to the looser definition [7].

SNP-protein and SNP-metabolite QTLs function as annotation layers, not as causal evidence

In the acute chest syndrome recurrence analysis, molecular QTL mapping produced 42 SNP-protein and eight SNP-metabolite associations at p<1e-3, and pathway and enrichment analysis over those hits identified secretory, platelet/hemostatic, and complement-mediated pathways, which the authors read as confirmation of a complex hemolytic-inflammatory pathophysiology [2]. The thresholds are the point: at p<1e-3 with no colocalization or mediation analysis reported, the molecular QTLs annotate the loci rather than resolve them, and the pathway output is a description of coherent biology attached to an already-defined locus set [2]. The dedicated sickle cell proteogenomic resource shows what a properly powered version of the SNP-protein arm looks like—5,411 plasma proteins in 343 patients yielding 560 pQTLs after conditional analysis, 58 of them novel—and it is still explicitly framed as a drug-target discovery exercise rather than as a mechanism for a specific recurrent phenotype [1].

Causal claims in these designs come from pQTL-instrumented randomization, not from enrichment

The paper that gets closest to mechanism combines its sickle cell pQTL map with genome-wide association results for fetal hemoglobin in a Mendelian randomization analysis and prioritizes five proteins (ENPP5, LBP, NAAA, PT3X, ZP3) as candidate HbF-increasing targets [1]. That instrumental-variable step is the one with external validation attached: pQTL-supported target-indication pairs were 4.7 times more likely to advance from Phase I to launch, against a 2.6-fold increase for human genetic evidence alone, with enrichment concentrated in druggable protein families that genetic evidence alone did not flag [8]. Population-scale atlases follow the same template, requiring both colocalization support and causal inference before calling a protein-phenotype pair high-confidence—645 such pairs in an East Asian proteogenomic atlas [6]—and combining GWAS, eQTL, pQTL and exome scores by minimum-rank integration gains roughly 558 further rank positions over GWAS alone across 30 traits [14].

The disease genotype contributes little new regulatory architecture, which caps what pathway enrichment can find

A finding that constrains the entire enterprise is that sickle cell pQTL effect sizes are largely concordant with those in non-sickle individuals from the Million Veteran Program, with heterogeneity evident only at specific loci such as APOL1 and haptoglobin [1]. If the causal mutation does not rewire the regulatory architecture of the plasma proteome, then SNP-protein mapping in a hemoglobinopathy cohort mostly rediscovers population pQTLs, and pathway enrichment over those hits describes shared physiology rather than disease-specific mechanism [1]. The heterogeneity at APOL1 and haptoglobin is the informative exception, and it is exactly where ancestry- and genotype-aware designs pay off: local ancestry-aware meta-analysis for sickle cell nephropathy identified ancestry-component-specific loci (12 regions in the African component, 38 in the European) mapping to inflammation, extracellular matrix, iron metabolism and magnesium homeostasis genes [4].

The metabolite arm captures enzyme-level biology the protein arm cannot, but is the least developed

Metabolite-side QTLs are the most underused component of these designs. Ratio QTL mapping across all pairwise metabolite comparisons identified 5,095 metabolite pairs with significant rQTLs, increased QTL count by 21% with 1,249 novel rQTLs across 53 loci, and showed the mapped genes are strongly enriched for enzymes—with 72% of rQTL-mapped genes unavailable in pQTL studies [5]. That complements the protein-centric view directly, and the SCD metabolomics literature independently identifies dyslipidemia and impaired oxidative metabolism in acute pain, with eight metabolic modules distinguishing healthy controls, baseline SCD, and acute pain states [12]. What does not yet exist in the retrieved literature is a study that genetically anchors those metabolomic modules in a hemoglobinopathy cohort; the acute chest syndrome work used only eight SNP-metabolite associations to feed its pathway analysis [2].

In other Mendelian settings, multi-omic QTLs are used to widen the phenome and explain penetrance rather than to map the primary locus

Where the causal gene is already known, integration is deployed downstream of it. Common-variant-based gene expression prediction for the Mendelian disease gene PLOD3 recovered the known BCARD phenome, added novel phenotypes that replicated in a PLOD3 pQTL analysis and identity-by-descent analysis, and generated new pathway-level associations when phenotypes were encoded as phenotype risk scores in a transcriptome-wide framework [10]. Multi-omic follow-up in an incompletely penetrant CHD8 family showed that despite similar haploinsufficiency, proband and unaffected mother had divergent downstream transcriptional effects, with methylation episignature positive in both—evidence that molecular phenotyping explains penetrance rather than assigns the locus [7]. In a sickle cell kidney model, single-cell multiomics combined with kidney pQTL data from genetically diverse mice showed Calb1 expression is highly heritable and co-regulated with Slc8a1, defining a heritable calcium-buffering axis that environmental iron status then modifies [17].

Conflicts and gaps

The retrieved papers disagree on whether molecular QTL annotation improves causal gene prioritization at all: benchmarking across 445 diseases found that neither eQTL colocalization nor locus-to-gene scores outperformed the simple nearest-gene method at predicting which genes became approved drug targets [13], in direct tension with the 4.7-fold pQTL enrichment reported elsewhere [8]. Colocalization yield and interpretation are also unsettled: 34–50% of GWAS hits colocalized in a large eQTL evaluation, over half in only one cell type, and 37–47% of colocalizing peaks did so with multiple genes [16], while method choice alone swings support for Bonferroni-significant protein-trait associations from 40.3% under coloc to 79.6% under SharePro [15]. A technical confound specific to Mendelian settings is that missense cis-pQTLs are more likely detected when the variant alters surface-exposed residues, i.e. when it interferes with reagent binding, so epitope effects contaminate a minority (≤12%) of cis-pQTL results precisely at the coding variants Mendelian disease cares about [9]. Two absences matter: no retrieved paper applies the SNP-protein plus SNP-metabolite plus enrichment pipeline to thalassemia or to a recurrent phenotype in any non-sickle hemoglobinopathy, and no retrieved paper reports formal mediation of a recurrent-phenotype association through a protein or metabolite intermediate in a hemoglobinopathy cohort.

What would change the answer

Region-based colocalization (coloc, coloc+SuSiE, PWCoCo, SharePro) between an acute chest syndrome recurrence GWAS and matched pQTL and mQTL data from the same cohort would convert the current p<1e-3 annotation into testable sharing of causal variants [15]. Applying metabolite ratio QTL mapping to sickle cell cohorts would add the enzyme-activity layer that pQTLs structurally miss [5], and mediation analysis linking SNP to protein or metabolite to recurrence would supply the arrow the current enrichment analyses leave undefined. Sample size is the binding constraint—343 patients in the largest SCD proteogenomic analysis [1]—so replication in larger multi-cohort consortia with ancestry-aware models is required [4]. Functional adjudication of prioritized proteins in patient-derived erythroid models is now feasible [3], and extending QTL mapping beyond abundance to phosphorylation and chromatin states would capture regulatory mechanisms invisible to pQTLs [11].

References

[1] Groza, C., Chignon, A., Lo, K. S., Bellegarde, V., Bartolucci, P., Lettre, G.. "Proteogenomic analysis of 5,411 plasma proteins in sickle cell disease patients". medRxiv. 2026-04-07T00:00:00.000Z. URL: https://doi.org/10.64898/2026.04.06.26350255

[2] Kulshrestha, A., Garrett, M. E., Nouraie, S. M., Machado, R. F., Gordeuk, V. R., Gladwin, M. T., Telen, M. J., Ashley-Koch, A. E., NHLBI Trans-Omics for Precision Medicine (TOPMed),. "Multi-omic and Survival Analysis of Acute Chest Syndrome Burden in Sickle Cell Disease". medRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.15.26365641

[3] Pate, B., Goldstein, A., Labott, M., Lizarralde-Iragorri, M., Chankhunthod, A., Tyson, T., Sloan, M., Wijeyesekera, C., Wilks, A., Steinberg, M. H., Murphy, G. J., Vanuytsel, K.. "Modeling Genetic Diversity in Sickle Cell Disease Reveals Heterogeneous Responses to HbF-Inducing Therapies". bioRxiv. 2026-05-21T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.18.726003

[4] Garrett, M. E., Nouraie, S. M., Machado, R. F., Gordeuk, V. R., Gladwin, M. T., NHLBI Trans-Omics for Precision Medicine Consortium,, Telen, M. J., Ashley-Koch, A. E.. "Local ancestry-aware genome-wide meta-analysis uncovers novel genetic loci for sickle cell disease nephropathy". medRxiv. 2026-05-30T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.27.26354213

[5] Rizi, S., Goss, N., Kutalik, Z., van der Graaf, A.. "Comprehensive metabolite ratio QTL mapping reveals disease relevant enzyme biology". medRxiv. 2025-12-05T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.04.25341616

[6] Pozarickij, A., Wang, B., Mohamed, A., Lin, K., Morris, S., Kartsonaki, C., Wright, N., Fry, H., Chen, Y., Du, H., Bennett, D., Yang, L., Avery, D., Schmidt, D. V., Lin, L., Lv, J., Yu, C., Sun, D., Pei, P., Chen, J., Hill, M., Peto, R., Collins, R., Clarke, R., Millwood, I. Y., Chen, Z., Walters, R. G.. "Genomic atlas of 7,000 plasma proteins and their associations with diseases and traits in East Asian populations". medRxiv. 2026-02-06T00:00:00.000Z. URL: https://doi.org/10.64898/2026.02.05.26345625

[7] Singh, M., Stenton, S. L., Groopman, E., Singer-Berk, M., Ma, J., Choufani, S., O'Heir, E., DiTroia, S., Osei-Owusu, I., Pais, L., Lemire, G., O'Leary, M., Austin-Tse, C., Genetti, C. A., Anderson, K. N., Wojcik, M. H., Beggs, A. H., Neil, J. E., Lai, A., Walsh, C. A., Jurgens, J. A., Barry, B. J., Engle, E. C., Donkervoort, S., Bonnemann, C. G., Argilli, E., Sherr, E. H., Kipkemoi, P., Bruwer, Z., Donald, K. A., Straub, V., Topf, A., White, S. M., Tan, T. Y., Gallacher, L., Arts, P., Jackson, M. R., Ha, T. T., Barnett, C. P., Byrne, A. B., Scott, H. S., Ganesh, V. S., Rehm, H. L., Weksberg, R. "Systematic evaluation of incomplete penetrance in 4,325 rare disease families reveals divergent transcriptional networks in CHD8 pathology". medRxiv. 2026-10-07T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.05.26364218

[8] Karim, M. A., Hukku, A., Ariano, B., Holzinger, E., Tsepilov, Y., Hayhurst, J., Buniello, A., McDonagh, E. M., Castel, S. E., Nelson, M. R., Maranville, J., Yerges-Armstrong, L., Ghoussaini, M.. "Impact of proteogenomic evidence on clinical success". medRxiv. 2026-03-05T00:00:00.000Z. URL: https://doi.org/10.64898/2026.02.23.26346731

[9] Kuliesius, J., Badonyi, M., Navarro, P., Marsh, J. A., Klaric, L., Wilson, J. F.. "Epitope Effect Prevalence in Affinity-based pQTL studies". bioRxiv. 2025-06-24T00:00:00.000Z. URL: https://doi.org/10.1101/2025.06.20.660695

[10] Scalici, A., Baker, J. T., Blostein, F., Shuey, M., Choudhary, D., Knapik, E. W., Samuels, D. C., Below, J. E., Bastarache, L., Miller-Fleming, T. W., Cox, N. J.. "Common variant approaches to study Mendelian disease gene function identify novel phenome and pathways associated with PLOD3". medRxiv. 2025-11-27T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.25.25340832

[11] Needham, E. J., Zhao, J. H., Hirschmuller, N., Wolford, C., Jiang, X., Butterworth, A. S.. "Genetic determinants of protein phosphorylation in humans". medRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.17.26365814

[12] Enders, J. D., Khalid, Z., Blecking, V., Ebert, A. D., Brandow, A. M., Stucky, C. L.. "Dyslipidemia is a metabolic hallmark of acute pain in sickle cell disease.". medRxiv. 2026-06-29T00:00:00.000Z. URL: https://doi.org/10.64898/2026.06.24.26356495

[13] Ji, C., Kerrebijn, I., Arbabi, K., Schipper, M., Wainberg, M.. "Benchmarking genome-wide association study causal gene prioritization for drug discovery". medRxiv. 2025-09-25T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.23.25336370

[14] Moix, S., Sadler, M. C., Kutalik, Z.. "Integration of genetic evidence to identify approved drug targets". bioRxiv. 2025-10-11T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.10.681636

[15] Zhang, W., Yoshiji, S., Sladek, R., Dupuis, J., Lu, T.. "Benchmarking Bayesian colocalization methods in validating Mendelian randomization-identified targets". bioRxiv. 2026-07-28T00:00:00.000Z. URL: https://doi.org/10.1101/2024.10.14.617627

[16] Reales, G., Pullin, J. M., Manipur, I., Vigorito, E., Wallace, C.. "Design and interpretation of eQTL-GWAS colocalisation studies: lessons from a large-scale evaluation". medRxiv. 2025-11-23T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.20.25340664

[17] AGORO, R., Xie, G., Janeczko, K., Williams, E., Wang, Y.-T.. "Genetic Variation, Iron Status, and FGF23 Signaling Converge to Regulate Renal Calcium Buffering in Sickle Cell Disease". bioRxiv. 2026-09-29T00:00:00.000Z. URL: https://doi.org/10.64898/2026.09.23.752598

How this survey was made

This survey answers one question: “How have multi-omic quantitative trait locus (QTL) integration approaches—such as SNP-protein and SNP-metabolite associations combined with pathway enrichment—been used to dissect genetic risk loci for severe recurrent phenotypes in hemoglobinopathies or other Mendelian diseases?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 17 papers whose work this text draws on, chosen from the 102 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 22, 2026 and covered work from the previous two years.