Table 3.

Eleven regions showing strong evidence of novel association $\left(\mathbb{P}\right({\mathbb{H}}_3\text{or}{\mathbb{H}}_4)>0.5)$ for an analysis involving a previously non–associated trait. D corresponds to T1D, R to RA, C to CEL and M to MS. Novel associations are underlined and denoted by bold font. Candidate causal genes are as associated across all curated diseases by ImmunoBase. Note that in the case of TNFAIP3, there is strong evidence that MS is caused by a distinct causal variant compared to the other traits. Distinct signals are separated by a ‘— ’. Since we only have a subset of the genotype data, not all of the prior (previously published) associations are seen.

Chromosome	Position	Prior Associations	Associations Found	Posterior probability both diseases are associated $\mathbb{P}\left({\mathbb{H}}_3\text{or}{\mathbb{H}}_4\right)$	Posterior probability shared causal variant given joint association $\mathbb{P}({\mathbb{H}}_4\mid {\mathbb{H}}_3\text{or}{\mathbb{H}}_4)$	Candidate Causal Genes/Genes in Region
1q24.3	170882016–171208336	C	DC	DC:0.75	DC:0.95	FASLG
2p14	65246601–65570598	R	RM	RM:0.86	RM:0.72	SPRED2
2q11.2	99883120–100415547	DR	DRC	DC:0.98 RC:1.00	DC:0.57 RC:0.90	AFF3
2q37.1	230758228–230962304	M	CM	CM:0.94	CM:0.90	SP140
5q11.2	55450712–55492884	RM	DRM	DR:0.71 DM:0.71	DR:1.00 DM:1.00	ANKRD55
6q23.3	137914792–138345363	RCM	DRC—M	DR:0.80 DC:0.77	DR:0.94 DC:0.93	TNFAIP3
7p14.2	37323488–37406978	CM	RCM	RC:0.80 RM:0.77	RC:0.84 RM:0.83	ELMO1
7p12.2	50222360–50335957	M	DM	DM:0.73	DM:0.70	5′ IKZF1* region
13q32.3	98723872–99034738	DM	D—C	DC:0.67	DC:0.00	GPR183
15q25.1	76773859–77050416	DM	DC	DC:0.82	DC:0.99	CTSH
19p13.2	10081000–11019034	DRM	DRM—C	DC:0.87 RC:0.87 CM:0.88	DC:0.40 RC:0.46 CM:0.57	ICAM1 ICAM3 TYK2

^*An association of T1D in a region 3′ of IKZF1, for which it is hypothesised that IKZF1 is the candidate causal gene is already known³³ (see Table 1). The novel association we report here is in a region 5′ of IKZF1, and independent of the established association. Note we provide coordinates of the region, and not an index SNP as is conventional in gwas studies because the method synthesises information across the whole region and does not, in most cases, highlight a single SNP responsible for the association.