Expanding carrier screening: beyond the genes, to include underrepresented ancestries

Abstract

Reproductive carrier screening has evolved beyond ethnic-specific testing to include diverse populations, yet gene selection varies considerably. In Singapore, genomic data analysis identified severe paediatric conditions amongst Chinese, Indian and Malay populations absent from existing screening panels. We developed a model leveraging data from 9051 participants to guide gene selection for carrier screening representative of Asian genetic diversity, focusing on severe paediatric-onset conditions prevalent in these populations. After evaluating severity, genotype-phenotype variability, clinical utility and technical feasibility, we identified 88 genes associated with recessive severe paediatric onset prevalent amongst Chinese, Indian and Malay populations, irrespective of carrier frequency. Including 24 additional genes from our registry resulted in a 105-gene panel, predicted to identify 0.44% at-risk couples, with 86 genes overlapping existing panels. Broadening criteria to include moderate severity conditions while limiting carrier frequencies to less than 1 in 200 reduced the panel to 59 genes, increasing predicted at-risk couples to 0.47%, due to higher carrier frequencies, yet introducing counselling complexities from greater clinical variability. Using local genomic data, we identified genetic conditions relevant to Asian populations for carrier screening. Expanding national genomic sequencing initiatives provides an opportunity to assess genetic condition prevalence across diverse ancestries, improving equity in carrier screening programmes.

Introduction

Genetic carrier screening offers information to reproductive couples about the likelihood of them having a child with a recessive genetic condition. The traditional approach of targeting at-risk populations, defined as an ethnic group based on geographic isolation or cultural identification for prevalent genetic conditions (such as Tay-Sachs disease in Ashkenazi Jewish and beta-thalassaemia in Southeast Asian populations¹) proved inequitable as it failed to recognise underrepresented populations or discordance with genetic ancestry^2–4. In response, professional guidelines evolved to recommend carrier screening to all couples planning a pregnancy⁵.

Genomic technologies enable multiple genes to be screened simultaneously, generating considerable debate regarding gene selection criteria⁶, such as disease severity, treatment availability, carrier frequency, and established genotype-phenotype associations. However, interpretation of these criteria varies across healthcare settings, particularly regarding assessment of disease severity and carrier frequency parameters. The American College of Medical Genetics and Genomics (ACMG) developed a practice resource which recommends screening genes with at least moderate severity and a carrier frequency of greater than 1 in 200 (0.5%) in any ethnic group². Using population data from the Genome Aggregation Database (gnomAD⁷) for six ancestries, classifications according to ClinVar⁸, and pre-defined disease severity^9,10, ACMG proposed 113 genes for screening. Although this approach improves equity across diverse populations, it is notable that the gnomAD database underrepresents certain ancestries, for example, those from Asian populations.

In Singapore, where 97% of the population is Chinese (74.3%), Malay (13.5%) and Indian (9.0%)¹¹, thalassaemia is the only condition for which reproductive carrier screening is offered routinely¹². However, an analysis of whole genome data from 9051 Singaporeans, comprising unrelated Chinese, Indian and Malay individuals, identified 10 genes described to be associated with severe paediatric recessive disease¹³, and with carrier frequencies exceeding 1 in 200, that were absent in commonly ordered commercial carrier screening panels¹⁴. This finding highlighted the need to expand carrier screening to include conditions prevalent in Asian populations and prompted further investigation. We used curated genomic data from the local Singaporean population and recommendations from professional societies to devise a gene list for carrier screening to increase equitable access for Asian populations, with consideration of what should be reported for reproductive planning purposes and how it could be implemented in Singapore. Using this gene panel, carrier screening will be offered to 1000 couples in the pilot phase and expanded to 39,000 couples across Singapore with no out-of-pocket costs. Reproductive risk will be reported for couples, and a genetic counselling appointment will be arranged for those at increased risk of having an affected pregnancy. During this appointment, test results and available reproduction options, such as prenatal testing and pre-implantation genetic testing (PGT) will be discussed. This study presents our framework for developing a gene panel optimised for Asian populations using a model which incorporates local genomic data.

Results

Selection of genes associated with severe paediatric conditions in the population

The cohort interrogated to characterise the carrier frequency of genetic conditions comprised 9051 individuals of Chinese (60.8%), Indian (21.4%) or Malay (17.8%) ancestry¹⁴. Variants in 993 genes of interest were identified in the cohort: pathogenic/ likely pathogenic (P/LP) variants were identified in 890 AR and 36 XL genes, and in a separate analysis, gross deletions in loss-of-function intolerant genes were detected in 67 autosomal recessive genes.

Guided by discussions with stakeholders and the limited experience in Singapore of genetic screening beyond thalassaemia, we focussed on severe conditions that present primarily in infancy or childhood. Referring to a published gene list curated for childhood onset¹³, 297 of the 993 genes with AR and XL inheritance carriers in the population cohort were associated with paediatric disease.

These genes were further interrogated to evaluate disease characteristics at both gene and variant levels, with clinical severity documented. While many genes are associated with clinical variability ranging from severe to mild phenotypes, we prioritised genes primarily associated with reduced lifespan in infancy or childhood and severe intellectual disability. This included conditions where severity could be distinguished by genotype. For example, only variant combinations associated with severe paediatric disease, such as CFTR and cystic fibrosis, rather than those associated with congenital bilateral absence of the vas deferens. A total of 97 genes met these criteria¹⁵.

Health implications for participants

Among the 97 genes, there were six genes associated with AD in addition to AR inheritance (ATM, BSCL2, COL4A3, COL4A4, GLA, LDLR and WFS1), and only one gene associated with XL inheritance. Given variants in these genes had implications for carriers, these were excluded.

There were also 18 cohort participants with homozygous variants in six genes associated with autosomal recessive conditions (CFTR, GALC, HBB, LAMA3, POLG, TUBGCP6) anticipated to result in a mild phenotype for homozygotes (Table 1). Although genes with implications for carriers were omitted, given their severity in a compound heterozygous state and prevalence in the population, these genes remained in the carrier screening panel.

Table 1.

The carrier screening panel consisting of 105 genes with gene-level carrier frequencies identified by ancestry group and number of homozygotes

Gene symbol	OMIM Phenotype	Condition type	Panel	Carrier frequency				No of homozgyotes	ACMG gene panel	Commercial screening panels1
				Chinese	Indian	Malay	Max carrier frequency#
ACAT1	Alpha-methylacetoacetic aciduria, 203750	Metabolic	NBS	0.0009	0.0010	0	0.0010		x	x
ADAR	Aicardi-Goutieres syndrome 6, 615010	Neurodegenerative	Severe NBS	0	0.0093	0	0.0093			x
AGT	Renal tubular dysgenesis, 267430	Renal	Severe NBS	0.0029	0.0005	0	0.0029			x
AGXT	Hyperoxaluria, primary, type 1, 259900	Metabolic	Severe NBS	0.0022	0.0015	0.0025	0.0025		x	x
AHI1	Joubert syndrome-3, 608629	Neurologic	Severe NBS	0.0022	0.0134	0.0019	0.0134		x	x
ALDH3A2	Sjogren-Larsson syndrome, 270200	Developmental	Severe NBS	0.0013	0.0010	0.0006	0.0013			x
ARSA	Metachromatic leukodystrophy, 250100	Lysosomal	Severe NBS	0.0022	0.0015	0.0031	0.0031		x	x
ASL	Argininosuccinic aciduria, 207900	Metabolic	NBS	0.0009	0.0021	0.0012	0.0021		x	x
ASPM	Microcephaly 5, 608716	Developmental	Severe NBS	0.0036	0.0021	0.0019	0.0036			x
BBS1	Bardet-Biedl syndrome 1, 209900	Developmental	Severe NBS	0.0004	0.0046	0.0025	0.0046		x	x
BBS2	Bardet-Biedl syndrome 2, 615981	Developmental	Severe NBS	0.0029	0.0005	0	0.0029		x	x
BCKDHA	Maple syrup urine disease, type Ia, 248600	Metabolic	NBS	0.0016	0.0005	0.0006	0.0016			x
BCKDHB	Maple syrup urine disease, type Ib, 248600	Metabolic	NBS	0.0002	0.0005	0.0012	0.0012		x	x
BLM	Bloom syndrome, 210900	Developmental	Severe NBS	0.0018	0.0026	0	0.0026		x	x
CC2D2A	Joubert syndrome 9, 612285	Neurologic	Severe NBS	0.0011	0.0010	0.0019	0.0019		x	x
CENPJ	Microcephaly 6, 608393	Developmental	Severe NBS	0.0018	0.0021	0	0.0021			x
CEP290	Joubert syndrome 5, 610188	Neurologic	Severe NBS	0.0109	0.0041	0.0031	0.0109		x	x
CFTR	Cystic fibrosis, 219700	Developmental	Severe NBS	0.0832	0.0350	0.1642	0.1642	11	x	x
CHRNE	Myasthenic syndrome, 605809	Neuromuscular	Severe NBS	0.0009	0.0015	0.0012	0.0015		x	x
CHRNG	Escobar syndrome, 265000	Developmental	Severe NBS	0.0009	0.0015	0.0012	0.0015			x
COLQ	Myasthenic syndrome, 603034	Neuromuscular	Severe NBS	0.0011	0.0015	0.0037	0.0037			x
COQ8B	Nephrotic syndrome, type 9, 615573	Renal	Severe NBS	0.0022	0	0	0.0022			x
CPS1	Carbamoylphosphate synthetase I deficiency, 237300	Metabolic	NBS	0.0005	0	0.0006	0.0006			x
CPT2	CPT II deficiency, lethal neonatal, 608836	Metabolic	NBS	0.0022	0	0.0006	0.0022		x	x
CYP17A1	17,20-lyase deficiency, isolated, 202110	Metabolic	NBS	0.0011	0	0	0.0011			x
CYP7B1	Bile acid synthesis defect, congenital, 3, 613812	Gastroenterologic	Severe NBS	0.0071	0.0005	0.0012	0.0071			x
DBT	Maple syrup urine disease, type II, 248600	Metabolic	NBS	0.0004	0.0015	0.0012	0.0015			x
DDC	Aromatic L-amino acid decarboxylase deficiency, 608643	Metabolic	Severe NBS	0.0062	0	0.0006	0.0062			x
DOK7	Myasthenic syndrome, congenital, 10, 254300	Neuromuscular	Severe NBS	0.0011	0.0026	0.0019	0.0026			x
DYM	Dyggve-Melchior-Clausen disease, 223800	Developmental	Severe NBS	0.0016	0.0015	0.0006	0.0016			x
DYNC2H1	Short-rib thoracic dysplasia 3, 613091	Skeletal	Severe NBS	0.0022	0.0021	0.0031	0.0031		x	x
EARS2	Combined oxidative phosphorylation deficiency 12, 614924	Mitochondrial	Severe NBS	0.0020	0.0005	0.0012	0.0020			x
EPG5	Vici syndrome, 242840	Developmental	Severe NBS	0.0011	0.0021	0.0006	0.0021			x
ERCC6	Cockayne syndrome, type B, 133540	Developmental	Severe NBS	0.0011	0.0021	0.0019	0.0021			x
EVC2	Ellis-van Creveld syndrome, 225500	Skeletal	Severe NBS	0.0018	0.0031	0.0006	0.0031		x	x
FANCA	Fanconi anaemia, complementation group A, 227650	Haematologic	Severe NBS	0.0040	0.0015	0.0006	0.0040			x
FKRP	Muscular dystrophy-dystroglycanopathy (congenital with brain and eye anomalies), type A, 613153	Neuromuscular	Severe NBS	0.0044	0.0015	0.0006	0.0044		x	x
FKTN	Muscular dystrophy-dystroglycanopathy (congenital with brain and eye anomalies), type A, 253800	Neuromuscular	Severe NBS	0.0022	0.0005	0.0012	0.0022		x	x
FRAS1	Fraser syndrome, 219000	Developmental	Severe NBS	0.0031	0.0021	0.0044	0.0044			x
GAA	Glycogen storage disease II, 232300	Lysosomal	Severe NBS	0.0145	0.0036	0.0006	0.0145		x	x
GALC	Krabbe disease, 245200	Lysosomal	Severe NBS	0.0160	0.0067	0.0031	0.0160	1		x
GCDH	Glutaricaciduria, type I, 231670	Metabolic	Severe NBS	0.0049	0.0021	0.0019	0.0049			x
GNPTAB	Mucolipidosis III alpha/beta, 252600	Lysosomal	Severe NBS	0.0025	0.0010	0.0019	0.0025		x	x
HBA1	Thalassaemias, alpha-, 604131	Haematologic	Severe NBS	0.0184	0.0005	0.0019	0.0184		x
HBA2	Thalassaemia, alpha-, 604131	Haematologic	Severe NBS	0	0	0			x
HBB	Thalassaemias, beta-, 613985	Haematologic	Severe NBS	0.0171	0.0118	0.0765	0.0765	2	x	x
HERC2	Mental retardation, 615516	Neurologic	Severe NBS	0.0018	0	0.0006	0.0018			x
HLCS	Holocarboxylase synthetase deficiency, 253270	Metabolic	Severe NBS	0.0036	0.0005	0	0.0036			x
HSD3B2	3-beta-hydroxysteroid dehydrogenase, type II, deficiency, 201810	Endocrine	NBS	0.0007	0	0	0.0007			x
IDUA	Mucopolysaccharidosis Ih, 607014	Lysosomal	Severe NBS	0.0013	0.0021	0	0.0021		x	x
INVS	Nephronophthisis 2, infantile, 602088	Renal	Severe NBS	0.0020	0	0	0.0020			x
IQCB1	Senior-Loken syndrome 5, 609254	Developmental	Severe NBS	0.0035	0	0	0.0035			x
KIAA0586	Short-rib thoracic dysplasia 14 with polydactyly, 616546	Skeletal	Severe NBS	0.0016	0.0010	0	0.0016			x
KLHL40	Nemaline myopathy 8, 615348	Neuromuscular	Severe NBS	0.0040	0.0005	0	0.0040			x
LAMA3	Epidermolysis bullosa, junctional, Herlitz type, 226700	Cutaneous	Severe NBS	0.0164	0.0026	0.0155	0.0164	1		x
LAMB3	Epidermolysis bullosa, junctional, Herlitz type, 226700	Cutaneous	Severe NBS	0.0011	0.0021	0.0006	0.0021			x
LINS1	Mental retardation, 614340	Neurologic	Severe NBS	0.0035	0	0	0.0035			x
MCPH1	Microcephaly 1, 251200	Developmental	Severe NBS	0.0053	0.0098	0.0118	0.0118		x	x
MKS1	Meckel syndrome 1, 249000	Developmental	Severe NBS	0.0031	0.0005	0.0012	0.0031			x
MMAA	Methylmalonic aciduria, 251100	Metabolic	NBS	0.0004	0.0005	0.0006	0.0006			x
MMAB	Methylmalonic aciduria, 251110	Metabolic	NBS	0.0002	0	0	0.0002		‘	x
MMACHC	Methylmalonic aciduria and homocystinuria, cblC type, 277400	Metabolic	NBS	0.0040	0.0015	0.0025	0.0040		x	x
MMADHC	Methylmalonic aciduria and homocystinuria, cblD type, 277410	Metabolic	NBS	0	0	0	0			x
MMUT	Methylmalonic aciduria, mut	Metabolic	Severe NBS	0.0031	0.0005	0.0019	0.0031		x	x
MYO5B	Microvillus inclusion disease, 251850	Gastroenterologic	Severe NBS	0.0016	0.0010	0	0.0016			x
NAGLU	Mucopolysaccharidosis type IIIB	Lysosomal	Severe NBS	0.0016	0.0015	0	0.0016			x
NEB	Nemaline myopathy 2, 256030	Neuromuscular	Severe NBS	0.0035	0.0052	0.0031	0.0052		x	x
NEU1	Sialidosis, type I, 256550	Lysosomal	Severe NBS	0.0020	0	0.0006	0.0020			x
NGLY1	Congenital disorder of deglycosylation, 615273	Metabolic	Severe NBS	0.0016	0.0005	0.0006	0.0016			x
NPC1	Niemann-Pick disease, type C1, 257220	Metabolic	Severe NBS	0.0022	0.0021	0.0019	0.0022			x
NPHP3	Meckel syndrome 7, 267010	Developmental	Severe NBS	0.0029	0.0010	0	0.0029			x
NPHS1	Nephrotic syndrome, type 1, 256300	Renal	Severe NBS	0.0031	0.0010	0.0006	0.0031		x	x
NPHS2	Nephrotic syndrome, type 2, 600995	Renal	Severe NBS	0.0018	0.0015	0.0025	0.0025			x
PAH	Phenylketonuria, 261600	Metabolic	Severe NBS	0.0098	0.0062	0.0025	0.0098		x	x
PKHD1	Polycystic kidney and hepatic disease, 263200	Developmental	Severe NBS	0.0056	0.0026	0.0031	0.0056		x	x
PLA2G6	Neurodegeneration with brain iron accumulation 2B, 610217	Neurodegenerative	Severe NBS	0.0022	0	0.0006	0.0022			x
PMM2	Congenital disorder of glycosylation, type Ia, 212065	Metabolic	Severe NBS	0.0058	0.0010	0.0006	0.0058		x	x
POLG	Mitochondrial DNA depletion syndrome, 203700	Mitochondrial	Severe NBS	0.0211	0.0026	0.0330	0.0330	2	x	x
POR	Antley-Bixler syndrome, 201750	Developmental	Severe NBS	0.0015	0.0005	0.0006	0.0015			x
PRDX1	Methylmalonic aciduria and homocystinuria, 277400	Metabolic	NBS	0	0	0	0
PTS	Hyperphenylalaninemia, BH4-deficient, A, 261640	Metabolic	Severe NBS	0.0038	0.0005	0	0.0038			x
QDPR	Hyperphenylalaninemia, BH4-deficient, C, 261630	Metabolic	NBS	0	0	0	0			x
RAD50	Nijmegen breakage syndrome-like disorder, 613078	Developmental	Severe NBS	0.0040	0.0031	0.0006	0.0040			x
RARS2	Pontocerebellar hypoplasia, type 6, 611523	Neurodegenerative	Severe NBS	0.0020	0.0010	0.0019	0.0020		x	x
RECQL4	Baller-Gerold syndrome, 218600	Skeletal	Severe NBS	0.0027	0.0021	0.0019	0.0027			x
RPGRIP1L	Meckel syndrome 5, 611561	Developmental	Severe NBS	0.0027	0.0005	0.0006	0.0027			x
SBDS	Shwachman-Diamond syndrome, 260400	Developmental	Severe NBS	0.0145	0.0067	0.0100	0.0145			x
SCN9A	Insensitivity to pain, congenital, 243000	Neurologic	Severe NBS	0.0035	0.0005	0	0.0035			x
SGSH	Mucopolysaccharidisis type IIIA, 252900	Lysosomal	Severe NBS	0.0011	0.0031	0.0019	0.0031			x
SLC25A20	Carnitine-acylcarnitine translocase deficiency, 212138	Metabolic	NBS	0.0024	0	0	0.0024
SLC26A2	Achondrogenesis Ib, 600972	Skeletal	Severe NBS	0.0013	0.0021	0	0.0021		x	x
SMN1	Spinal muscular atrophy-1, 253300	Neuromuscular	Severe NBS	0	0	0	0.0192		x	x
SPINK5	Netherton syndrome, 256500	Cutaneous	Severe NBS	0.0055	0.0015	0	0.0055			x
SPR	Dystonia, dopa-responsive, 612716	Neurologic	Severe NBS	0	0	0.0006	0.0006			x
STAR	Lipoid adrenal hyperplasia, 201710	Endocrine	NBS	0.0009	0	0.0012	0.0012			x
TBC1D24	Epileptic encephalopathy, early infantile, 16, 615338	Neurologic	Severe NBS	0.0033	0	0.0006	0.0033			x
TCIRG1	Osteopetrosis, 259700	Skeletal	Severe NBS	0.0024	0.0010	0	0.0024			x
TH	Segawa syndrome, 605407	Neurologic	Severe NBS	0.0040	0	0.0037	0.0040			x
TMEM237	Joubert syndrome 14, 614424	Neurologic	Severe NBS	0.0013	0	0.0025	0.0025			x
TTC21B	Short-rib thoracic dysplasia 4, 613819	Skeletal	Severe NBS	0.0005	0.0021	0.0025	0.0025			x
TTC7A	Gastrointestinal defects and immunodeficiency syndrome, 243150	Gastroenterologic	Severe NBS	0.0016	0	0.0006	0.0016			x
TUBGCP6	Microcephaly and chorioretinopathy, 251270	Developmental	Severe NBS	0.0022	0.0015	0	0.0022	1		x
VPS13B	Cohen syndrome, 216550	Developmental	Severe NBS	0.0042	0.0021	0.0012	0.0042			x
WRN	Werner syndrome, 277700	Developmental	Severe NBS	0.0049	0	0.0019	0.0049			x
ZFYVE26	Spastic paraplegia 15, 270700	Neuromuscular	Severe NBS	0.0027	0.0015	0	0.0027			x
Gene symbol	OMIM Phenotype	Condition type	Panel	Carrier frequency				No of homozgyotes	ACMG gene panel	Commercial screening panels
				Chinese	Indian	Malay	Max carrier frequency#
HBA1, HBA2	Thalassaemias, alpha-, 604131	Haematologic	Severe	0.01836	0.00052	0.00187	0.00187	0	x	x
SMN1	Spinal muscular atrophy-1, 253300	Neuromuscular	Severe				0.01920	0	x	x

Severe severe paediatric phenotype, NBS newborn screening, RDF rare disease fund, PGT pre-implantation genetic testing. #The highest carrier frequency among the three ancestry groups. *Severe severe paediatric condition, NBS newborn screening.

¹Commercial panels referenced for comparison: Myriad Foresight (176 genes), Centrogene CentoScreen® Solo (330 genes), Fulgent Beacon (787 genes), Invitae carrier screening panel (556 genes).

Technical limitations

Two genes deemed technically challenging, CLCNKB and CYP21A2, due to the presence of homologous pseudogenes¹⁶, were also excluded.

Local screening initiatives

Using the Singaporean population data, the resulting panel included 88 genes prevalent in at least one of the three predominant ancestries associated with autosomal recessive inheritance and severe paediatric disease. We also took into consideration 46 genes included in newborn screening (NBS) in Singapore. Of these, 17 ranged in clinical variability, which poses challenges for reproductive decision making; for example, BTD associated with biotinidase deficiency. Three X-linked genes were excluded due to implications for carriers: IL2RG and OTC, and GCH1. Two genes, CYP11B1 and CYP21A2, were technically difficult to sequence. There was consensus that 24 conditions were of clinical value for carrier screening. Seven were already included due to their severity, so an additional 17 were added to the carrier screening panel.

Gene panel for implementation

The resulting panel included 105 genes for carrier screening implementation, comprising 88 genes relevant to Chinese, Malay and Indian populations (Fig. 1) and further 17 genes from newborn screening. Although presence amongst the predominant ancestries was a selection criterion, carrier frequency thresholds were not applied. As a result, carrier frequencies ranged widely, from HBB (7% in the Malay population) to ALDH3A2 (0.1% in the Chinese population). Using a panel of this size helped lower sequencing costs by allowing multiplexing of up to 192 samples in a single sequencing run, which was a central consideration during design, given that genetic testing is not typically subsidised or covered by insurance in Singapore. Additionally, the selection of genes was designed to minimise health implications for carriers, thereby streamlining resources for pre and post-test genetic counselling, and potentially reducing barriers to participation arising from personal health concerns.

Fig. 1.

Development of a carrier screening panel for Chinese, Indian and Malay populations comprising of 105 genes.

The development of this carrier screening panel involved consideration of multiple factors, including population-specific genetic prevalence, age of onset, severity and prognosis of genetic conditions as well as consideration to the cost of development. The health implications for heterozygous carriers and the integration with existing screening programmes, such as newborn screening, also played a role in determining which genetic conditions should be included in the screening panel. With these considerations in mind, we mapped our preferences according to gene selection variables and outcomes, which defined the panel size. This model can be adapted with the emergence of new data or priorities (Fig. 2).

Fig. 2.

A model demonstrating the interaction of gene selection variables and outcomes mapped according to panel size. The priorities of this screening programme are outlined in orange, which focused on an affordable yet ancestry-inclusive gene panel to facilitate implementation and genetic counselling delivery. As such, the conditions selected for screening were present in the population with variable frequencies, characterised by severe phenotypes and confined to autosomal recessive inheritance.

Relevance to the Singaporean population

For this carrier screening initiative, only at-risk couples where both partners are found to carry a P/LP variant in the same gene associated with a severe paediatric phenotype will be reported. The frequency of couples considered at risk of having an affected pregnancy associated with severe disease in the 105 gene panel was calculated. Only at-risk couples of the same genetic ancestry were considered, as they have been shown to confer a higher carrier risk compared to admixed couples¹⁰. Variation across ancestries was observed, with couples of Malay ancestry having the highest risk (0.44%), followed by Chinese (0.30%) and Indian (0.17%). Given there are close to 30,000 births per year in Singapore, this equates to 23 affected pregnancies. With the inclusion of copy number variants detected in alpha thalassaemia (HBA1/HBA2) and spinal muscular atrophy (SMN1) that have overall carrier frequencies of 1.16% and 1.92% respectively¹⁰, 27 births per year are estimated to be affected based on the devised 105 gene panel (Table 2). For comparison, referring to the SG10K_Health cohort, there were 2888 participants found to be a carriers of at least one condition screened in the 105 gene panel, resulting in 36% of the cohort carrying a P/LP variant.

Table 2.

. At-risk couples and affected pregnancies amongst Chinese, Indian and Malay

Carrier couple frequency %
	Genes (n)
	105	59
Chinese	0.30	0.43
Indian	0.17	0.35
Malay	0.44	0.47
HBA1/2 CNV	1.16	1.16
SMA CNV	1.92	1.92

Births affected (n) (30,000 births/yr)
	Genes (n)
	105	59
Chinese	17	25
Indian	1	2
Malay	5	5
HBA1/2 CNV	1	1
SMA CNV	2	2
Total	27	36

Comparison to existing carrier screening panels

Comparing the 105 gene list to existing available carrier screening panels, only 35 of the 105 genes overlapped with genes recommended by ACMG for carrier screening² and of those excluded, 7 genes had a carrier frequency of >1 in 200 with severe paediatric onset (ADAR, CYP7B1, DDC, GALC, LAMA3, SBDS and SPINK5). Further interrogation identified several variables to account for the differences. First, we referenced a broader panel of Mendelian paediatric disorders (1246 genes¹³ versus ACMG 924 gene list⁹), accounting for the exclusion of ADAR, CYP7B1 and SPINK5 from the ACMG carrier screening panel. Second, variant curation methods differed for LAMA3 as ACMG used ClinVar P/LP variants, regardless of the submission gold star rating status and potentially inflating carrier frequency estimates¹⁰, thereby impacting overall prevalence. Finally, disease prevalences varied between the two cohorts sharing similar ancestries. For example, the carrier frequency of DDC was 0.6% in our Chinese cohort but was absent amongst East Asians in the gnomAD curated data used by ACMG¹⁰. Comparing our 105 gene panel to commercial providers approved in Singapore, there was an overlap of 86 genes (Table 1).

The development of a carrier screening panel using alternative selection criteria

Adopting a pragmatic approach to align with stakeholder preferences and existing resources in our local setting, including preferencing genes that had severe paediatric onset, our proposed carrier screening panel diverged from ACMG recommendations, which includes conditions with at least moderate severity, XL inheritance and a prevalence of >1 in 200 in any ancestry. To explore alternate outcomes, we postulated a gene panel based on recommendations by ACMG using the genomic analysis of our local ancestry-specific data. There were 37 genes with a prevalence of >1 in 200 in either Chinese, Indian or Malay associated with at least moderate severity, referring to the most recent curated list of 1246 genes for clinical utility^13,17 and of these, 18 are not present in the ACMG 113 carrier screening gene panel.

The addition of 22 genes relevant to newborn screening resulted in a panel of 59 genes (Supplementary Table 1). Although smaller, the number of at-risk couples increased across all ancestries, Chinese 0.43%, Indian 0.35% and Malay 0.47%, identifying 36 affected pregnancies per year (Table 2). This increase is attributed to the addition of conditions with a higher prevalence yet variability in clinical presentation. There were 19 conditions with a prevalence of >1 in 100 with moderate severity, in comparison to the 105 gene panel which has 12 conditions with a prevalence of at least 1 in 100. For example, SLC25A13 (Citrin deficiency, OMIM 605814) and ATP7B (Wilson disease, OMIM 277900) have a maximum carrier frequency of 1 in 44 and 1 in 56 consecutively, however they are associated with variable onset of symptoms and severity ranging from child to adulthood^18,19. These genes were not included in our gene panel as they may pose complexities in counselling and decision making due to their variable expressivity.

Discussion

We developed a reproductive screening panel tailored to the Singaporean population, taking into account Asian diversity. Large public genomic databases, such as gnomAD, provide carrier frequencies across ancestries, which can be referenced for the purpose of carrier screening^2,10. Likewise, genetic testing companies have estimated prevalence according to ancestry based on uptake of their screening tests^15,20. However, we noted differences between the carrier frequencies of the Singapore population and the representation of Asian ancestry in gnomAD, likely due to the different origins of the populations. For example, individuals of Malay ancestry are absent in gnomAD and the ancestries comprising East Asian are not well defined, impacting the inclusion of genes for carrier screening. Furthermore, some variation may be attributed to ancestry documentation subject to self-reported bias or curation processes relying on ClinVar classification entries^10,14. These considerations underscore the limitations of relying solely on established databases for diverse populations. In this study, we leveraged genomic data available from Singapore population characterised by genetic ancestry. Using population data interrogated by manually curated genomic variants amongst Chinese, Indian and Malay individuals, we could design a gene panel that was relevant to the targeted population. We also incorporated genes included in the newborn screening programme. This approach could also consider other genetic conditions prioritised for clinical care initiatives, such as those receiving subsidised treatment funding or approved by the Ministry of Health for preimplantation genetic testing.

Cost was a priority for implementing carrier screening in Singapore, and extends to Southeast Asia, given that genetic testing is generally out-of-pocket and can present barriers to access²¹. Minimising the cost of the panel can enhance accessibility and also increase the likelihood of government funding or subsidies. The consideration of clinical severity largely shapes inclusion and the number of genes to screen. Given the experience of carrier screening in Singapore has centred around thalassaemia, the preferences and acceptability for further inclusion are unknown. Insights from other countries implementing national screening programmes indicate support for screening serious conditions associated with significant intellectual disability and reduced lifespan in childhood^22,23. In this context, reproductive interventions can be considered acceptable and less challenging in terms of decision-making for family planning purposes^24,25. Consequently, conditions with severe health implications were prioritised, reducing the number of genes for screening and therefore the cost of panel development and delivery.

Health implications for carriers were also considered, as these require substantial resources in pre and post-test counselling to address insurance implications and individual screening recommendations. For these reasons, we removed recessive and X-linked genes that have implications for heterozygous carriers. In practice, those who would prefer the inclusion of X-linked conditions can be referred to the genetics clinic, where a commercial panel is available at an additional cost. Although rare, homozygotes were also detected in this study cohort, highlighting that individual health risks and reproductive implications should still be considered during implementation. At an individual level, the implementation of this screening panel will identify approximately 36% of tested individuals as carriers. However, as the personal health implications to recessive heterozygous carriers are limited and delivery of individual results is resource intensive^26,27 and can result in increased anxiety^27,28, simultaneous screening and reporting of reproductive risk pertaining to the couple is preferred.

We did not restrict carrier frequency as a selection criteria, and consequently, only 20 genes had carrier frequencies exceeding 1 in 200. Refining the criteria as guided by ACMG² to include moderately severe conditions but exclude conditions with a carrier frequency of less than 1 in 200 would yield a smaller gene panel of 59 genes while increasing the detection of at-risk couples. However, using this model includes conditions associated with variable clinical outcomes or later onset (e.g., GNE-myopathy, OMIM 600737), which may not alter reproductive choices²⁹. Overall, selecting genes for carrier screening involves balancing the benefits of identifying severe conditions with the risks of uncertain clinical outcomes, which can challenge reproductive decision making²⁵. The interpretation of variants associated with incomplete penetrance or variable expressivity requires extensive curation and genetic counselling resources. While commercial carrier screening companies demonstrate the feasibility of screening genes with variable severity, and there is perceived value from consumers^24,30, additional clinical genetics resources are required for this approach, which may fall outside the scope of a national screening programme.

Although this panel was developed based on the data available at the time of review, we acknowledge that gaps in the available information may exist. The prevalence of genetic conditions was determined by the presence of P/LP variants amongst Chinese, Indian and Malay amongst a cohort of 9051 participants. Although variants were curated using up-to-date classification resources, causative variants for all conditions will not have been observed within this cohort size. This cohort is currently increasing to 100,000 participants in Phase II of Singapore’s National Precision Medicine programme (www.npm.sg), which can be further interrogated to refine the prevalence of recessive conditions amongst the Singapore population. The genes were selected for this panel as they are associated with shortened lifespan or severe intellectual disability. Only variants associated with severe disease will be reported, however, we acknowledge that severity can be challenging to infer. Furthermore, disease severity was taken predominantly from the perspective of stakeholders, yet can be influenced by a range of factors such as lived experience and available support services. To date, most studies have been conducted in healthcare settings with European-derived populations, and further research is required regarding participation and barriers amongst diverse populations to understand acceptability and impact³¹. The inclusion of genes for carrier screening can be modified over time to respond to community expectations and revised knowledge regarding condition prevalence and severity.

The expansion of nationwide genomic sequencing initiatives worldwide provides an opportunity to characterise the prevalence of genetic conditions in diverse ancestries. We outline key variables that implicate gene selection and panel development, and demonstrate how these will vary based on the intended outcomes of the carrier screening programme, while accounting for the targeted population, available healthcare resources and local priorities. Irrespective of inclusion criteria, the genes selected for carrier screening should be population appropriate and balanced with clinical benefit to achieve the fundamental goal of informed reproductive decision-making.

Methods

Gene selection criteria

Our primary aim was to develop a carrier screening panel that is inclusive of the Singaporean population. This involved selecting genetic conditions prevalent among the three predominant ancestries - Chinese, Indian and Malay. As equitable access was a priority, the cost of implementation was also considered, with focus on the number of genes included, the technical complexity of screening these genes and processes for returning results. The development of our carrier screening panel was guided by key criteria derived from literature and existing recommendations. There is consensus that phenotype severity and treatment availability should be considered, as these can impact reproductive decisions. The prevalence of genetic conditions in the targeted population, analytical validity of the screening methods and established genotype-phenotype associations have also been included in gene selection criteria^2,4,32. However, the application of these criteria for gene panel selection varies considerably according to assessment of condition severity, carrier frequency and existing screening initiatives in the local healthcare setting, as summarised in Table 3. To address these complexities, we conducted weekly meetings from November 2022 to September 2024 with a multidisciplinary team comprising of clinical geneticists, genetic counsellors, genetics nurses, bioinformaticians and molecular genetics scientists to discuss in detail the clinical utility and technical feasibility of gene selection.

Table 3.

Gene selection considerations for carrier screening

Gene selection assessment
Criteria	Society recommendations	Considerations
Relevance to the target population	Relevant to all ancestries1	•Prevalence data may be unavailable for all ancestries within the target population •Carrier frequency threshold such as 1 in 1002, 1 in 2001
Age of onset	Paediatric onset3	Variable age of onset
Disease severity	At least moderate severity1	Variable in severity presentation – profound, severe, moderate, mild
Genotype-phenotype correlation	Established genotype-phenotype association1–3	Genes associated with multiple phenotypes ranging from well-characterised to less defined
Personal health risks	Resources to support health risks to carriers1,3,4	Manifestations of symptoms for heterozygote carriers associated with XL1 and AD conditions or as homozygous or compound heterozgyotes for AR conditions

Relevance to local context
Cost: funded/ subsidised/ out of pocket costs
Technical assay ability
Clinical utility -treatment available or reproductive management options
Established screening programmes e.g. newborn screening

¹ Gregg, A.R. et al., Screening for autosomal recessive and X-linked conditions during pregnancy and preconception: a practice resource of the American College of Medical Genetics and Genomics (ACMG). Genet Med, 2021. 23(10): p. 1793-1806.

²Committee Opinion No. 690. Carrier screening in the age of genomic medicine. Obstet. Gynecol. 129, e35–e40 (2017).

³Henneman, L. et al., Responsible implementation of expanded carrier screening. Eur J Hum Genet, 2016. 24(6): p. e1-e12.

⁴Edwards, J.G. et al., Expanded carrier screening in reproductive medicine-points to consider: a joint statement of the American College of Medical Genetics and Genomics, American College of Obstetricians and Gynaecologists, National Society of Genetic Counsellors, Perinatal Quality Foundation, and Society for Maternal-Foetal Medicine. Obstet Gynecol, 2015. 125(3): p. 653-662.

Relevance to the target population

The genomic data interrogated to characterise the carrier frequency of genetic conditions amongst the Singaporean population is derived from Singapore’s National Precision Medicine Programme known as SG10K_Health. This reference database comprised 10,000 genomes from participants with no reported pre-existing medical conditions. Of these, kinship analysis identified 9051 participants unrelated to the second degree and genetic ancestry was inferred using ADMIXTURE³³. For the purpose of this analysis, ‘ancestry’ refers to the genetic ancestry of Singapore’s major population groups: Chinese, Indian, and Malay. Detailed methods regarding the generation of this cohort as well as whole genome sequencing methods, bioinformatic analysis and variant curation have been previously described in the SG10K_Health project¹⁴. Single nucleotide, insertions, deletions and copy number variants amongst the study cohort of 9051 participants occurring in 4143 genes associated with autosomal dominant (AD), autosomal recessive (AR) and X-linked (XL) Mendelian disease were classified according to ACMG guidelines.

Using the data generated from SG10K_Health, the frequency of each pathogenic/likely pathogenic (P/LP) variant in the population was established, and the proportion of participants who carry a P/LP variant as well as the overall carrier rate for each gene according to genetic ancestry, Chinese, Indian and Malay, was determined for AR and XL conditions as these are relevant for carrier screening. The carrier frequency for each gene was documented; however, thresholds were not applied as selection criteria for the gene panel, with the exception of copy number variant screening.

Ethics approval and consent from participants contributing to whole genome data from a cohort of 9051 Singaporeans was obtained from six institutes: Growing Up in Singapore Towards healthy Outcomes birth cohort study (GUSTO), National University Hospital, Singapore CIRB/E/2019/2655 NCT01174875; Health for Life In Singapore Study (HELIOS), Nanyang Technology University, Singapore, IRB 2016-11-030; Singapore Multi-Ethnic Cohort Study (MEC), National University of Singapore, CIRB 13-512; SingHealth Duke-NUS Institute of Precision Medicine (PRISM), SingHealth Centralised Institutional Review Board, 2013/605/C NCT02791152; Singapore Epidemiology of Eye Diseases study (SEED), SingHealth Centralised Institutional Review Board, 2018/2717; Tan Tock Seng Hospital (TTSH), National Health Group, TB-2020-001 & BTC-2020-001. All experiments were performed in accordance with relevant guidelines and regulations. This research conforms to the principles of the Helsinki Declaration.

Age of onset

To identify genes associated primarily with paediatric onset we referred to a published gene list curated for childhood onset, life limiting or disabling conditions comprising of 1246 genes^13,17.

Clinical severity and utility

Severity was assessed in relation to impact on life span, intellect and mobility, in conjunction with variability in age of onset, penetrance and expressivity¹⁵, using both published literature and clinical experience. Assessment also included a review of treatment availability and associated costs, family planning options, the number of cases observed at the tertiary paediatric hospital, historic prenatal testing requests, as well as international screening practices and views. The clinical genetics team comprised four clinical geneticists with expertise in paediatric conditions, two genetic counsellors and one genetic nurse. Each condition was independently assessed by one of the clinical geneticists and reviewed by the genetic counsellors and one genetic nurse. Any differing opinions were resolved through the team discussion. When necessary, additional input was sought regarding technical development and stakeholder review was also employed. Conditions were defined as severe if they were associated with reduced lifespan in childhood and/ or severe to profound intellectual disability.

Genotype-phenotype correlation

Gene disease validity was reviewed to reflect emerging evidence from sources such as ClinVar⁸, ClinGen³⁴, OMIM³⁵ and published literature. For genes associated with multiple conditions and inheritance patterns, each P/LP variant was further interrogated using the same sources to establish the clinical significance at a variant level.

Health risks to gene carriers

In Singapore, genetic testing in asymptomatic individuals, which may infer increased risk of a medical condition can have insurance implications and must be clearly communicated in pretest counselling. We evaluated the health implications for carriers recognising the significance to genetic counselling resources and insurance implications. After deliberation, variants posing health risks for heterozygotes associated with recessive, X-linked and dominant conditions were excluded as we did not want concerns regarding personal health implications to deter participation. This approach aligned with the overarching aim of the programme which was to support inclusion and reproductive decision-making.

Cost of testing

Drawing on the experience of newborn screening in Singapore, which has an out-of-pocket cost of USD 150 and an uptake rate of 92% (unpublished data), this was used as a reference point when considering cost. This is similar to costs for carrier screening reported in other countries, ranging from USD 109 in China³⁶ to USD 266 in the Netherlands³⁷.

Technical assay ability

The analytical ability to detect clinically relevant variants by the laboratory was assessed and validated. Technical challenges impacting sequencing such as homologous or repeat regions within genes and the presence of pseudogenes, were taken into account¹⁶. An independent assay was developed to detect copy number variants for clinically relevant genes that were highly prevalent in the population, such as HBA1/2 and SMN1.

Established screening programmes in the local setting

In addition to considering genetic conditions prevalent in the population, we evaluated genetic conditions included in newborn screening, comprising 46 genes. For each gene, the number of cases detected in our tertiary hospital, its clinical utility for reproductive planning and views towards carrier screening for these conditions published in the literature were documented independently by four clinical geneticists with expertise in newborn screening. Discrepancies were subsequently resolved by discussion until consensus was achieved.

Stakeholder engagement/ perspectives

A short list of genes with corresponding genetic conditions and carrier frequencies were reviewed with a group of stakeholders that included the obstetrics and gynaecology department at a tertiary hospital, government representatives, funding organisations and consumer representatives. We also engaged with religious leaders (representatives from Islam, Buddhism, Christianity, Hinduism, Sikhism and Judaism) to share the carrier screening programme aims and conditions to be screened. An overview of the gene evaluation process is provided in Fig. 3.

Fig. 3.

Overview of gene selection review process for the carrier screening panel development.

Estimation of at-risk couples and affected births per year

The frequency of couples at risk of having a child associated with the genetic condition considered for carrier screening was estimated using an algorithm previously described¹⁴.^38, All simulated couples, irrespective of ancestry, were determined to be at risk if both carried a P/LP variant associated with severe disease in the same gene. Simulated couples with variants in the same gene associated with a mild phenotype or hypomorphic in the homozygous state were excluded.

The frequency of gross deletions occurring in loss-of-function intolerant genes was investigated in a separate analysis. The number of affected births was calculated based on 30,000 live births per year in Singapore³⁹.

Author contributions

Conceptualization: S.J., W.K., P.C.J., Y.B. Data curation: S.J., Y.B., W.K., J.T. Formal analysis: S.J., S.J., W.K., P.C.J., Y.B., J.T., S.L., R.W. Funding acquisition: S.J., W.K. Investigation: S.J., W.K., P.C.J., Y.B., J.G., S.L., R.W., S.L, J.T. Methodology: S.J., J.G., Y.B, C.C., D.A., J.H., M.M. Writing review and editing: Y.B., S.J., D.A., W.K., P.C.J., J.G., C.C., S.L, J.H, M.M, RW. All authors read and approved the final manuscript.

Data availability

The datasets analysed during the current study are available from the corresponding author on reasonable request.

Competing interests

SSJ is the co-founder of Global Gene Corp Pte Ltd and Rhea Health Pte Ltd. SSJ has received travel allowances from Illumina, Pacific BioSciences and Oxford Nanopore. DA is the Medical Director at Eugene Labs. WK is the co-founder of Rhea Health Pte Ltd. The rest of the authors declare that they have no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41525-025-00545-w.