Power of Inclusion: enhancing polygenic prediction with admixed individuals

Tanigawa and Kellis. Am J Hum Genet. (2023).

Phenotype: Mean sphered cell volume

Link to the source data field in UK Biobank
Estimated h² in white British population in UKB: 0.166 (95% CI:[0.147, 0.185]).
Link to PGS catalog (Study ID: PGP000502).
GWAS Catalog for mapped trait(s):

Mean sphered cell vol. iPGS coefficients

Our FAQ page shows the description of the file format and how you may use iPGS coefficients in your research.

iPGS prediction in the held-out test set individuals

We compared the polygenic prediction from our iPGS model and the phenotype values using the held-out test set individuals in UK Biobank. Note the difference in the number of individuals in the five population groups.

/static/data/tanigawakellis2023/per_trait/INI30270/INI30270.WB.PGS_vs_phe.png

/static/data/tanigawakellis2023/per_trait/INI30270/INI30270.NBW.PGS_vs_phe.png

/static/data/tanigawakellis2023/per_trait/INI30270/INI30270.SA.PGS_vs_phe.png

/static/data/tanigawakellis2023/per_trait/INI30270/INI30270.Afr.PGS_vs_phe.png

/static/data/tanigawakellis2023/per_trait/INI30270/INI30270.others.PGS_vs_phe.png

Predictive performance

Population	Model	Metric	Predictive Performance	95% CI	P-value
Population	Model	Metric	Predictive Performance	95% CI	P-value

white British	Covariate-only model	R²	0.012	[0.011, 0.014]	2.7x10^-175
white British	Genotype-only model	R²	0.153	[0.148, 0.158]	<1.0x10^-300
white British	Full model (covariates and genotypes)	R²	0.164	[0.159, 0.169]	<1.0x10^-300
Non-British white	Covariate-only model	R²	0.024	[0.013, 0.035]	2.1x10^-16
Non-British white	Genotype-only model	R²	0.155	[0.131, 0.180]	1.2x10^-103
Non-British white	Full model (covariates and genotypes)	R²	0.179	[0.154, 0.204]	1.8x10^-120
South Asian	Covariate-only model	R²	0.017	[0.004, 0.030]	1.2x10^-06
South Asian	Genotype-only model	R²	0.103	[0.074, 0.132]	1.2x10^-34
South Asian	Full model (covariates and genotypes)	R²	0.116	[0.086, 0.147]	3.7x10^-39
African	Covariate-only model	R²	0.002	[-0.003, 0.007]	1.5x10^-01
African	Genotype-only model	R²	0.033	[0.013, 0.052]	9.4x10^-10
African	Full model (covariates and genotypes)	R²	0.034	[0.014, 0.054]	5.7x10^-10
Others	Covariate-only model	R²	0.039	[0.030, 0.047]	6.8x10^-67
Others	Genotype-only model	R²	0.137	[0.123, 0.151]	6.6x10^-245
Others	Full model (covariates and genotypes)	R²	0.164	[0.149, 0.178]	7.3x10^-296

The predictive performance (R²), its 95% confidence interval (CI), and statistical significance (P-value) are shown for each population in UK Biobank in the held-out test set. The "model" column indicates whether the predictive performance is from the covariate-terms alone (covariate-only model), PGS terms alone (Genotype-only model), or the full model containing both PGS and covariate terms. We used the following sets of covariates in our analysis: age, sex, age², age*sex, Townsend deprivation index, and genotype PCs (PC1-PC18). Please refer to our publication for a more detailed description of the methods.

Coefficients (BETA) of PGS models

/static/data/tanigawakellis2023/per_trait/INI30270/INI30270.BETAs.png

We show the coefficients (BETA) of PGS models. Our iPGS model selected 21558 variants with non-zero coefficients. The genetic variants with the large absolute values of coefficients are annotated in the plot. There is no guarantee that our iPGS model selects causal variants. We use the GRCh37/hg19 reference genome.

The top 100 genetic variants with the largest absolute value of coefficients

CHROM	POS	Variant	Variant ID	Effect allele	Consequence	Gene symbol	Beta
CHROM	POS	Variant	Variant ID	Effect allele	Consequence	Gene symbol	Beta

1	158637728	1:158637728:T:C	rs148912436	C	PAVs	SPTA1	-1.233
8	41543675	8:41543675:G:A	rs34664882	A	PAVs	ANK1	-0.948
8	41630405	8:41630405:G:A	rs4737009	A	Intronic	ANK1	0.635
X	153763492	X:153763492:T:C	rs1050829	C	PAVs	G6PD	0.519
6	139842599	6:139842599:G:T	rs653513	T	Others		-0.471
6	139840693	6:139840693:A:C	rs592423	C	Others		-0.423
4	55394172	4:55394172:C:T	rs218237	T	Others		0.389
7	50428445	7:50428445:T:C	rs12718598	C	Intronic	IKZF1	0.377
6	25918225	6:25918225:T:C	rs80215559	C	Intronic	SLC17A2	0.377
19	33754548	19:33754548:C:T	rs78744187	T	Others		-0.328
14	65267469	14:65267469:T:C	rs230703	C	PAVs	SPTB	0.305
1	158580069	1:158580069:C:T	rs2479868	T	Others	SPTA1, OR10Z1	0.299
1	158577109	1:158577109:A:C	rs857685	C	PAVs	OR10Z1	0.291
12	112331317	12:112331317:G:A	rs12580246	A	PTVs	MAPKAPK5	0.280
6	42010420	6:42010420:C:G	rs12200388	G	Intronic	CCND3	-0.273
6	26091179	6:26091179:C:G	rs1799945	G	PAVs	HFE	0.256
1	248039451	1:248039451:C:T	rs3811444	T	PAVs	TRIM58	-0.251
22	37469590	22:37469590:C:T	rs387907018	T	PAVs	TMPRSS6	0.234
6	135419018	6:135419018:T:C	rs9399137	C	Intronic	HBS1L	0.229
3	142295368	3:142295368:C:A	rs6782400	A	Intronic	ATR	-0.229
20	55989808	20:55989808:C:T	rs99595	T	Others		0.224
4	146055117	4:146055117:T:C	rs11545157	C	UTR	OTUD4	-0.210
2	8750266	2:8750266:A:G	rs3856447	G	Intronic	AC011747.6	0.204
11	100456604	11:100456604:C:T	rs11224302	T	Others		-0.204
10	71099888	10:71099888:G:A	rs10159477	A	Intronic	HK1	0.202
1	25818182	1:25818182:T:C	rs115340137	C	Intronic	TMEM57	-0.200
8	19813529	8:19813529:A:G	rs268	G	PAVs	LPL	-0.199
9	4856877	9:4856877:G:A	rs10758658	A	Intronic	RCL1	-0.198
1	203273366	1:203273366:A:G	rs4971234	G	Intronic	LINC01136	0.196
6	41925159	6:41925159:G:A	rs9349205	A	Intronic	CCND3	-0.196
17	76121864	17:76121864:A:G	rs2748427	G	PAVs	TMC6	-0.194
6	41877671	6:41877671:G:A	rs114056237	A	Intronic	MED20	0.191
12	4330215	12:4330215:C:T	rs4573764	T	Others		0.189
19	12999458	19:12999458:C:T	rs8110787	T	Others	KLF1, GCDH	0.179
9	100740124	9:100740124:C:T	rs4743150	T	Others		-0.178
22	50971266	22:50971266:T:C	rs140522	C	Others	TYMP, ODF3B	0.176
8	41663896	8:41663896:A:G	rs10103618	G	Intronic	ANK1	0.172
18	43833701	18:43833701:T:TCTG	rs34068795	TCTG	PAVs	C18orf25	-0.172
1	158580477	1:158580477:A:G	rs12128171	G	Others	SPTA1, OR10Z1	-0.171
2	112167931	2:112167931:T:C	rs62160676	C	Intronic	MIR4435-1HG	0.169
1	25572984	1:25572984:T:G	rs34484514	G	PAVs	C1orf63	0.164
12	57146069	12:57146069:T:G	rs2277339	G	PAVs	PRIM1	0.160
1	25570081	1:25570081:T:C	rs1043879	C	PAVs	C1orf63	0.158
8	41566438	8:41566438:C:T	rs2304877	T	PAVs	ANK1	-0.157
1	43415346	1:43415346:G:T	rs710220	T	Intronic	SLC2A1	0.150
6	41905275	6:41905275:G:A	rs3218097	A	Intronic	CCND3	-0.145
9	115950114	9:115950114:C:T	rs45559933	T	PAVs	FKBP15	-0.145
6	42017286	6:42017286:T:C	rs12210681	C	Intronic	CCND3	0.144
12	54685880	12:54685880:C:T	rs35979828	T	Intronic	RP11-968A15.8	0.144
6	41925304	6:41925304:G:A	rs11968166	A	Intronic	CCND3	0.144
7	123411223	7:123411223:A:C	rs4731120	C	Others		-0.142
7	100240296	7:100240296:A:G	rs2075672	G	Intronic	TFR2	0.141
16	87886490	16:87886490:C:T	rs68149176	T	Intronic	SLC7A5	0.141
17	4901982	17:4901982:C:T	rs11658587	T	Intronic	KIF1C	-0.141
9	91401893	9:91401893:C:T	rs9410344	T	Others		-0.139
6	41921241	6:41921241:G:T	rs10947997	T	Intronic	CCND3	0.138
12	53757831	12:53757831:A:G	rs12582170	G	Others		0.136
1	25583610	1:25583610:C:G	rs72660908	G	Intronic	C1orf63	-0.134
6	135431318	6:135431318:T:C	rs6920211	C	Others		0.134
6	31506691	6:31506691:G:A	rs2071596	A	PAVs	DDX39B	-0.133
19	17252151	19:17252151:T:C	rs35365035	C	Intronic	MYO9B	0.133
19	1056492	19:1056492:G:C	rs3752246	C	PAVs	ABCA7	-0.132
6	135381351	6:135381351:A:G	rs11759077	G	PTVs	CTA-212D2.2	-0.131
17	42300278	17:42300278:C:T	rs10208	T	Others	RP5-882C2.2	-0.130
9	91569248	9:91569248:G:T	rs11137467	T	Others		-0.129
17	44061023	17:44061023:G:A	rs62063786	A	PAVs	MAPT	-0.129
6	41848061	6:41848061:C:T	rs34718512	T	Intronic	USP49	-0.127
3	24350811	3:24350811:A:G	rs9310736	G	Intronic	THRB	0.127
19	4366219	19:4366219:A:G	rs732716	G	Intronic	SH3GL1	-0.125
6	41925290	6:41925290:T:A	rs11970772	A	Intronic	CCND3	0.123
12	21331549	12:21331549:T:C	rs4149056	C	PAVs	SLCO1B1	-0.121
16	30103160	16:30103160:C:A	rs3809627	A	UTR	TBX6	-0.120
6	41915519	6:41915519:G:A	rs6934551	A	Intronic	CCND3	0.120
2	112278539	2:112278539:G:A	rs61033544	A	Others		0.120
8	19819724	8:19819724:C:G	rs328	G	PTVs	LPL	0.120
6	109323519	6:109323519:G:T	rs2273668	T	PAVs	SESN1	-0.120
12	120955578	12:120955578:C:A	rs11065147	A	Intronic	COQ5	-0.118
1	198545552	1:198545552:T:G	rs1080976	G	Others		0.118
1	3691997	1:3691997:AGTCAGCCTAGGGGCTGT:A	rs566629828	A	PTVs	SMIM1	0.117
11	9769562	11:9769562:C:G	rs415895	G	PAVs	SWAP70	-0.117
8	145112983	8:145112983:C:T	rs55916375	T	PAVs	OPLAH	0.117
17	16852187	17:16852187:A:G	rs34557412	G	PAVs	TNFRSF13B	-0.115
2	174219135	2:174219135:G:A	rs920112	A	Intronic	AC092573.2	0.115
7	99269397	7:99269397:T:C	rs6977165	C	PTVs	CYP3A5	0.114
19	45422946	19:45422946:A:G	rs4420638	G	Others	APOC1	-0.114
16	67219107	16:67219107:G:C	rs9939768	C	PAVs	EXOC3L1	0.113
2	27730940	2:27730940:T:C	rs1260326	C	PAVs	GCKR	0.113
1	20915701	1:20915701:A:C	rs2072671	C	PAVs	CDA	0.111
1	203301965	1:203301965:C:A	rs12040268	A	Others		-0.111
11	94886632	11:94886632:T:C	rs496321	C	Intronic	RP11-712B9.2	-0.110
22	21982892	22:21982892:C:T	rs2298428	T	PAVs	YDJC	-0.110
4	55394323	4:55394323:A:G	rs10488853	G	Others		0.109
8	41507237	8:41507237:C:T	rs7825337	T	Intronic	NKX6-3	0.108
6	41681183	6:41681183:G:A	rs17794619	A	Intronic	TFEB	0.107
19	4409756	19:4409756:A:G	rs2230636	G	PAVs	CHAF1A	-0.107
3	195921311	3:195921311:G:A	rs9325434	A	Others	ZDHHC19	-0.107
12	121163518	12:121163518:C:A	rs2239760	A	Others	RP11-173P15.5, ACADS	-0.106
14	65249815	14:65249815:C:T	rs6573568	T	Intronic	SPTB	-0.104
18	60920854	18:60920854:C:T	rs17758695	T	Intronic	BCL2	0.103
7	50417177	7:50417177:T:C	rs118050676	C	Intronic	IKZF1	-0.103

There is no guarantee that our iPGS model selects causal variants. We show the top 100 variants with the largest effect size (BETA). To see 21558 variants included in our iPGS model, please download the iPGS coefficients by clicking the download button. We use the GRCh37/hg19 reference genome.

Follow-up analysis

There are several ways to use the resource in your research. First, you may use our iPGS coefficients and compute individual-level polygenic scores for your cohort. Second, you may also investigate the genetic variants with non-zero coefficients and their annotated genes to learn more about biology by taking advantage of the sparsity of our iPGS models. For your convenience, here we suggest several resources as an example of follow-up analysis. We do not intend to cover all the relevant follow-up analyses.

Using iPGS coefficients

By clicking the download button above, you may download the iPGS coefficients. Our FAQ page shows the description of file format and how you may use iPGS coefficients in your research.

HaploReg

HaploReg is a tool for exploring annotations of the non-coding genome at variants on haplotype blocks. The button above submits the top 100 genetic variants with the largest absolute value of coefficients as a query to HaploReg using the default parameters in HaploReg v4.2 (LD threshold r² >= 1, ChromHMM 15-state model, SiPhy-omega, and GENCODE genes). HaploReg's ability to browse haplotypes is useful here as there is no guarantee that our iPGS model selects causal variants. The 'top 100 variant' cutoff is an arbitrary threshold; we aim to demonstrate how one may investigate the selected variants. Please check Ward and Kellis. Nucleic Acids Res. 2012 and Ward and Kellis. Nucleic Acids Res. 2016 for more information on HaploReg.

GREAT

GREAT: Genomic Regions Enrichment of Annotations Tool evaluates enrichment of pathway and ontology terms. The ability of GREAT to map non-coding genetic variants to their downstream target genes would be suitable for investigating pathway and ontology enrichment of genetic variants selected in our sparse iPGS model. The button above submits the top 1000 genetic variants with the largest absolute value of coefficients as a query to GREAT using the default parameters in GREAT v4.0.4. The 'top 1000 variant' cutoff is an arbitrary threshold; we aim to demonstrate how one may investigate the selected variants. Please check McLean et al. Nat Biotechnol. 2010 and Tanigawa*, Dyer*, and Bejerano. PLoS Comput Biol. 2022 for more information on GREAT.

References

Y. Tanigawa and M. Kellis. Power of inclusion: Enhancing polygenic prediction with admixed individuals. Am J Hum Genet. (2023).
- We introduce our inclusive PGS approach in this publication.
- Supplementary Data Files at figshare (doi: 10.6084/m9.figshare.22905368)
- MIT News article: Making genetic prediction models more inclusive
- Author interview: Inside AJHG: A Chat with Yosuke Tanigawa
J. Qian, Y. Tanigawa, W. Du, M. Aguirre, C. Chang, R. Tibshirani, M. A. Rivas, T. Hastie. A fast and scalable framework for large-scale and ultrahigh-dimensional sparse regression with application to the UK Biobank. PLoS Genet. 16, e1009141 (2020).
- In iPGS, we fit penalized multivariate regression on individual-level data using the BASIL algorithm implemented in the R snpnet package. This publication describes the BASIL algorithm and the R snpnet package.