*2.5. MTNR1B Gene Score*

Genomic DNA was extracted from EDTA-buffered whole blood using Puragene ™ DNA Isolation Kit (Gentra Systems, Plymouth, UK). Genotyping of SAPALDIA participants was done on DNA samples collected at SAP2. In the framework of the EU-funded GABRIEL consortium to identify genetic determinants of asthma [35], 1612 participants were genotyped using Illumina Human 610Kquad BeadChip (G1; Illumina, San Diego, CA, USA) covering ~570,000 variants. An additional 3015 participants were genotyped using Illumina Human OmniExpress-Exome BeadChip (G2; Illumina, San Diego, CA, USA), covering ~1 million variants. Quality control criteria were applied to both genotyping arrays: samples with <97% genotyping success rate, or of non-European origin, with cryptic relatedness or sex inconsistencies, were excluded. Variants with minor allele frequency (MAF) of <5% or deviation from Hardy–Weinberg equilibrium (HWE) at a threshold of 10−<sup>6</sup> were also excluded. G1 and G2 datasets were phased using ShapeIT version 2.r790 [36] and imputed using MiniMac2 [37]. The imputed datasets were then merged, after excluding variants with low imputation quality (R<sup>2</sup> < 0.3), yielding ~14 million markers for 4324 participants across G1 and G2, from where we

identified seven common *MTNR1B* variants (rs1387153, rs10830962, rs4753426, rs8192552, rs10830963, rs3781638 and rs2166706) involved in glucose dysregulation [21]. Imputations were of high quality. All but rs10830963 (R<sup>2</sup> ≥ 0.87) had an imputation R<sup>2</sup> ≥ 0.92. All variants had similar allele frequencies in comparison with other studies [38–49] and the 1000 Genomes Central European Population [50]. All variants were in HWE (*p* > 0.2), with MAF ≥ 7% (Table S1). Two variants, rs10830962 and rs2166706, were in high LD (R<sup>2</sup> = 0.89; Figure S2), thus, we excluded rs2166706 from the analyses. The LD R<sup>2</sup> of the included MTNR1B variants ranged between 0.2 and 0.7. Since the glucose-raising allele of rs10830963 was the allele implicated in melatonin profile dysregulation, expressed as a significant delay in melatonin offset [22], we coded the other variants in their reported direction of association in glucose alterations, such that each variant contained 0–2 quantities of the risk allele. We created MGRS by summing up risk alleles across the six variants, yielding a minimum, median, and maximum of 2, 6, and 12 risk alleles, respectively. We also created a categorical variable—low risk (MGRS ≤ 6) and high risk (>6).
