共定位

步骤 5 · 措辞升级关卡 · 图 3 张 · 发表版表格 2 个　（流程见 v5 定稿）

不通过者保留在结果里报告，只是不升级措辞。MHC 区照常算、照常报 PP.H4，但该值不足以单独升级。

图

Figure F5a. Colocalisation between plasma cis-pQTL signals and each diabetic complication GWAS, for the 52 protein-outcome pairs (28 proteins) that reached step 5

完整图注（英文，与投稿版一致）

Figure F5a. Colocalisation between plasma cis-pQTL signals and each diabetic complication GWAS, for the 52 protein-outcome pairs (28 proteins) that reached step 5.

Method. coloc.abf (Giambartolomei et al. 2014, PLoS Genetics 10:e1004383) applied to a 1 Mb window (+/- 500 kb) centred on the cis gene, with the default single-causal-variant assumption and priors p1 = p2 = 1e-4, p12 = 1e-5. Cells give the posterior probability of a shared causal variant (PP.H4) taken directly from coloc_verdict.csv; values below 0.01 are shown in scientific notation because PP.H4 spans 286 orders of magnitude across this set. Tiers: strong PP.H4 >= 0.8, moderate 0.5 to 0.8, weak below 0.5.

Role in the analysis pipeline. Step 5 is a wording gate, not a veto. A pair that does not colocalise is retained in every downstream table; only its permitted wording is restricted. 13 pairs (11 proteins) meet the criterion for upgraded wording, defined as strong evidence outside the MHC region. Cross-disease sign consistency is reported separately and does not enter this decision.

The MHC region is shown separately for a reason. All 29 MHC-region pairs fall in the weak or moderate tier and 28 of them have PP.H3 = 1.00, meaning the data favour two distinct causal variants rather than one shared variant. This is the expected behaviour of a single-causal-variant method in a region of extended linkage disequilibrium and high gene density, and it should not be read as evidence against these proteins. Establishing or excluding colocalisation here requires MHC-specific work not performed in this study: conditional analysis, HLA allele-level colocalisation, and cross-platform replication. Two of these proteins are of direct pharmacological interest, complement factor B (the target of iptacopan) and AGER, so their position in the weak tier is a limitation of the method in this region, not a negative result.

Empty cells. A blank cell means the pair did not reach step 5, either because the Mendelian randomisation association was not significant at step 2 or because the pair was vetoed earlier. Of the 112 protein-by-outcome combinations shown, 52 were analysed and 60 were not. A blank cell is therefore not a negative colocalisation result.

Complication endpoints from FinnGen R9 (general-population controls); type 1 diabetes GCST90824163; type 2 diabetes Mahajan et al. 2018 (non-UK Biobank). Exposures: cis-pQTLs for 1,954 plasma proteins from UK Biobank Pharma Proteomics Project (Olink Explore; discovery cohort, n = 34,557 European ancestry).

Data sources and sample sizes. Exposure: UKB-PPP plasma proteome (Olink Explore 3072), 34,557 European participants, 1,954 proteins with a cis-pQTL. Outcomes: FinnGen R9 -- Diabetic retinopathy 10,413 cases / 308,633 controls; Diabetic maculopathy 3,572 cases / 308,547 controls; Diabetic nephropathy 4,111 cases / 308,539 controls; Diabetic neuropathy 2,843 cases / 271,817 controls.

完整图注（英文，与投稿版一致）

Figure F5b. Outcome sharing among the 11 proteins whose plasma cis-pQTL colocalises strongly (PP.H4 >= 0.8) with at least one diabetic complication.

What is plotted. Only the 13 pairs in the strong tier are included, so that the figure shows the set of results whose wording is upgraded and nothing else. Vertical bars count proteins per outcome combination; the dot matrix below identifies the combination; horizontal bars give the number of proteins per outcome. 9 of the 11 proteins colocalise with a single complication and 2 colocalise with two.

Diabetic neuropathy is shown with a set size of zero. No protein reaches the strong tier for this outcome. Neuropathy was analysed at step 5 on the same footing as the other three complications (5 pairs entered), so this is a measured negative result and not missing data. Its bar is retained at zero rather than dropped, so that the distinction between an outcome that was tested and returned nothing and an outcome that was never tested remains visible.

No MHC-region protein appears here. All 11 MHC-region proteins that reached step 5 fall in the weak or moderate tier, so none qualifies for the strong tier and the figure needs no MHC stratification. See Figure F5a and its caption for why the single-causal-variant model underperforms in that region.

Method: coloc.abf on a 1 Mb cis window, priors p1 = p2 = 1e-4, p12 = 1e-5. Colocalisation is a wording gate and does not remove candidates; proteins absent from this figure remain in all downstream tables.

Complication endpoints from FinnGen R9 (general-population controls); type 1 diabetes GCST90824163; type 2 diabetes Mahajan et al. 2018 (non-UK Biobank).

完整图注（英文，与投稿版一致）

Figure F5d. Outcome sharing among the 19 proteins whose plasma cis-pQTL colocalises strongly (PP.H4 >= 0.8) with at least one of the four complications or the two diabetes control arms, after removing every association vetoed earlier in the workflow.

Relation to Figure F5b. F5b shows the four complications only, which is the set whose wording may be upgraded. This figure adds the two diabetes control arms on the same footing, so that colocalisation with the complication can be read against colocalisation with diabetes liability itself. Restricted to the complications, the two figures agree exactly: 13 pairs and 11 proteins.

Reading it. 8 proteins colocalise with a diabetes control arm but with NO complication. Together with the 61% of associations classified as diabetes-driven in Figure F-DM-b, this is colocalisation-level support for the same conclusion: much of the plasma-protein signal at these loci acts on liability to diabetes rather than on the complication.

Exclusions. Associations vetoed at step 3 (reverse Mendelian randomization) or step 4X (epitope and protein-altering-variant investigation) are removed; this matters most for the control arms, where reverse causation is common. No MHC filter is applied because none is used anywhere in this workflow, and in any case no MHC-region protein reaches the strong tier for any of the six outcomes. Diabetic neuropathy is retained at zero: it was analysed and returned nothing.

Method: coloc.abf (Giambartolomei 2014) on a 1 Mb cis window, priors p1 = p2 = 1e-4 and p12 = 1e-5, strong tier PP.H4 >= 0.8.

发表版表格

文件	下载
`colocalisation_all_pairs.csv`	下载
`colocalisation_verdict.csv`	下载

本页图为网页版 PNG（最宽 1600 px）。投稿用的矢量 PDF 与 600 dpi TIFF 体积较大，留在仓库 results/coloc/figures/，不随文档站分发。

共定位 ​

图 ​

Figure F5a. Colocalisation between plasma cis-pQTL signals and each diabetic complication GWAS, for the 52 protein-outcome pairs (28 proteins) that reached step 5 ​

Figure F5b. Outcome sharing among the 11 proteins whose plasma cis-pQTL colocalises strongly (PP.H4 >= 0.8) with at least one diabetic complication ​

Figure F5d. Outcome sharing among the 19 proteins whose plasma cis-pQTL colocalises strongly (PP.H4 >= 0.8) with at least one of the four complications or the two diabetes control arms, after removing every association vetoed earlier in the workflow ​

发表版表格 ​

共定位

图

Figure F5a. Colocalisation between plasma cis-pQTL signals and each diabetic complication GWAS, for the 52 protein-outcome pairs (28 proteins) that reached step 5

Figure F5b. Outcome sharing among the 11 proteins whose plasma cis-pQTL colocalises strongly (PP.H4 >= 0.8) with at least one diabetic complication

Figure F5d. Outcome sharing among the 19 proteins whose plasma cis-pQTL colocalises strongly (PP.H4 >= 0.8) with at least one of the four complications or the two diabetes control arms, after removing every association vetoed earlier in the workflow

发表版表格