Kati Hyvärinen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Kati Hyvärinen has a total of 195 co-authors. Here are the top co-authors by number of shared publications:
- Jukka Partanen
Blood Service Biobank, Finnish Red Cross Blood Service, Vantaa, Finland
16 shared publications - Pirkko J. Pussinen
Department of Oral and Maxillofacial Diseases University of Helsinki Helsinki Finland
14 shared publications - Jarmo Ritari
Research and Development Finnish Red Cross Blood Service Helsinki Finland
13 shared publications - Susanna Paju
Department of Oral and Maxillofacial Diseases University of Helsinki Helsinki Finland
7 shared publications Eija Könönen
Institute of Dentistry University of Turku Turku Finland
5 shared publicationsSaara Laitinen
Finnish Red Cross Blood Service, Helsinki, Finland
5 shared publications
Latest Publications
- Open Access
Replication of recipient genetic variants associated with survival after lung transplantation
Background: Long-term survival after lung transplantation (LTx) remains limited, largely because of chronic lung allograft dysfunction and other post-transplant complications. Genetic studies have suggested recipient single-nucleotide polymorphisms (SNPs) to associate with LTx outcomes, but none has been confirmed in independent cohorts. Methods: We evaluated 29 previously published candidate SNPs associated with LTx complications in 201 LTx recipients from FinnGen. Time-to-event analyses were performed for death after transplantation, with censoring at end of follow-up. Each SNP was tested under additive, dominant, and recessive genetic models using Cox proportional hazards regression adjusted for sampling age, sampling year, sex, and the first ten genetic principal components. Replication was defined as nominal association with survival at p < 0.05, surpassing minimum detectable effect at 80% power and concordant direction of effect with the original publication. Results: We found two independent replicated signals: rs4526713-T and rs4149117-T. In multivariable analysis, rs4526713-T and rs4149117-T were associated with increased mortality risk under an additive and dominant models, respectively. A two-SNP score showed a stronger association with mortality than any individual SNP, suggesting an aggregate contribution of replicated genetic risk alleles to post-transplant survival. Conclusions: Our findings support the contribution of host genetic variation to long-term LTx outcomes. However, modest cohort size and use of all-cause mortality rather than allograft dysfunction-specific outcomes require cautious interpretation. Larger donor-recipient genomic studies with harmonized transplant phenotypes are needed before clinical implementation.
Open AccessReplication of recipient genetic variants associated with survival after lung transplantation
Background: Long-term survival after lung transplantation (LTx) remains limited, largely because of chronic lung allograft dysfunction and other post-transplant complications. Genetic studies have suggested recipient single-nucleotide polymorphisms (SNPs) to associate with LTx outcomes, but none has been confirmed in independent cohorts. Methods: We evaluated 29 previously published candidate SNPs associated with LTx complications in 201 LTx recipients from FinnGen. Time-to-event analyses were performed for death after transplantation, with censoring at end of follow-up. Each SNP was tested under additive, dominant, and recessive genetic models using Cox proportional hazards regression adjusted for sampling age, sampling year, sex, and the first ten genetic principal components. Replication was defined as nominal association with survival at p < 0.05, surpassing minimum detectable effect at 80% power and concordant direction of effect with the original publication. Results: We found two independent replicated signals: rs4526713-T and rs4149117-T. In multivariable analysis, rs4526713-T and rs4149117-T were associated with increased mortality risk under an additive and dominant models, respectively. A two-SNP score showed a stronger association with mortality than any individual SNP, suggesting an aggregate contribution of replicated genetic risk alleles to post-transplant survival. Conclusions: Our findings support the contribution of host genetic variation to long-term LTx outcomes. However, modest cohort size and use of all-cause mortality rather than allograft dysfunction-specific outcomes require cautious interpretation. Larger donor-recipient genomic studies with harmonized transplant phenotypes are needed before clinical implementation.
- Open Access
- 18 October 2026
- Published by Wiley in British Journal of Haematology
Identification of rare blood group donors from biobank data
Summary: Safe blood transfusion relies on avoiding clinically significant incompatibilities between recipient antibodies and donor red blood cell antigens, a requirement that becomes particularly challenging for patients with rare blood group phenotypes, especially when alloimmunized. Conventional approaches to identifying rare blood group donors rely on targeted screening and laborious laboratory studies, which are limited by resource constraints and insufficient population diversity. The aim of this study was to enhance the identification of rare blood group donors in Finland by utilizing genome‐wide data from our national Blood Service Biobank. Genomic variants associated with rare phenotypes were screened in 56 997 blood donors, after which the rare blood group was confirmed using routine methods. To estimate the need for rare blood group units, we utilized data from the Helsinki University Hospital (HUS) Data Lake and the laboratory information system of the HUS Blood Bank. Eighty‐six new rare donors were identified, with complete concordance between genomic results and confirmed phenotypes. Patient data highlighted donor needs and identified rare phenotypes, particularly those more prevalent among individuals of African ancestry. These findings demonstrate the value of biobank genomic data in expanding rare donor registries and improving transfusion preparedness.
- 18 October 2026
- Published by Wiley in British Journal of Haematology
Open AccessIdentification of rare blood group donors from biobank data
Summary: Safe blood transfusion relies on avoiding clinically significant incompatibilities between recipient antibodies and donor red blood cell antigens, a requirement that becomes particularly challenging for patients with rare blood group phenotypes, especially when alloimmunized. Conventional approaches to identifying rare blood group donors rely on targeted screening and laborious laboratory studies, which are limited by resource constraints and insufficient population diversity. The aim of this study was to enhance the identification of rare blood group donors in Finland by utilizing genome‐wide data from our national Blood Service Biobank. Genomic variants associated with rare phenotypes were screened in 56 997 blood donors, after which the rare blood group was confirmed using routine methods. To estimate the need for rare blood group units, we utilized data from the Helsinki University Hospital (HUS) Data Lake and the laboratory information system of the HUS Blood Bank. Eighty‐six new rare donors were identified, with complete concordance between genomic results and confirmed phenotypes. Patient data highlighted donor needs and identified rare phenotypes, particularly those more prevalent among individuals of African ancestry. These findings demonstrate the value of biobank genomic data in expanding rare donor registries and improving transfusion preparedness.
- Open Access
- 23 April 2026
- Published by American Society for Clinical Investigation in JCI Insight
- Vol. 11 (11)
Pan-cancer polygenic risk score associates with cancer susceptibility following kidney transplantation
BACKGROUND: Cancer accounts for over 20% of late posttransplant mortality, yet the contribution of genetic susceptibility to posttransplant cancer risk remains unclear. This study investigates germline genetic risk factors for posttransplant cancer in the Finnish population using data from the FinnGen cohort. METHODS: A pan-cancer polygenic risk score (PRS) was constructed using genetic variants identified in UK and US populations to assess the influence of common germline variants on time to first cancer diagnosis in 1,802 Finnish kidney transplant recipients (KTRs), of whom 317 developed posttransplant cancer. The PRS was first validated in the FinnGen non-transplantation cohort and subsequently applied to KTRs, with replication in lung and liver transplant recipients ( n = 476). Functional relevance was explored by assessing associations between the PRS and expression levels of 2,923 plasma proteins in the UK Biobank ( n = 53,013). RESULTS: Compared with a matched non-transplantation cohort ( n = 68,294), KTRs exhibited earlier cancer onset. The PRS was significantly associated with time to first cancer diagnosis in the non-transplantation population (HR 1.04, 95% CI 1.038–1.056, P = 3.75 × 10 –25 ). Among KTRs younger than 40 years, higher PRS was associated with earlier cancer onset (HR 1.08, 95% CI 1.01–1.17, P = 0.036), indicating a stronger genetic effect at younger ages. The PRS significantly (Bonferroni < 0.05) altered the regulation of 87 plasma proteins, several of which were known cancer-related markers. CONCLUSION: Inherited genetic predisposition, captured by pan-cancer PRS, may contribute to individual susceptibility to cancer after solid organ transplantation, particularly at younger ages. FUNDING: State research funding (Helsinki and Uusimaa Health District), the Foundation for Pediatric Research, and the Sigrid Jusèlius Foundation.
- 23 April 2026
- Published by American Society for Clinical Investigation in JCI Insight
- Vol. 11 (11)
Open AccessPan-cancer polygenic risk score associates with cancer susceptibility following kidney transplantation
BACKGROUND: Cancer accounts for over 20% of late posttransplant mortality, yet the contribution of genetic susceptibility to posttransplant cancer risk remains unclear. This study investigates germline genetic risk factors for posttransplant cancer in the Finnish population using data from the FinnGen cohort. METHODS: A pan-cancer polygenic risk score (PRS) was constructed using genetic variants identified in UK and US populations to assess the influence of common germline variants on time to first cancer diagnosis in 1,802 Finnish kidney transplant recipients (KTRs), of whom 317 developed posttransplant cancer. The PRS was first validated in the FinnGen non-transplantation cohort and subsequently applied to KTRs, with replication in lung and liver transplant recipients ( n = 476). Functional relevance was explored by assessing associations between the PRS and expression levels of 2,923 plasma proteins in the UK Biobank ( n = 53,013). RESULTS: Compared with a matched non-transplantation cohort ( n = 68,294), KTRs exhibited earlier cancer onset. The PRS was significantly associated with time to first cancer diagnosis in the non-transplantation population (HR 1.04, 95% CI 1.038–1.056, P = 3.75 × 10 –25 ). Among KTRs younger than 40 years, higher PRS was associated with earlier cancer onset (HR 1.08, 95% CI 1.01–1.17, P = 0.036), indicating a stronger genetic effect at younger ages. The PRS significantly (Bonferroni < 0.05) altered the regulation of 87 plasma proteins, several of which were known cancer-related markers. CONCLUSION: Inherited genetic predisposition, captured by pan-cancer PRS, may contribute to individual susceptibility to cancer after solid organ transplantation, particularly at younger ages. FUNDING: State research funding (Helsinki and Uusimaa Health District), the Foundation for Pediatric Research, and the Sigrid Jusèlius Foundation.
- Open Access
- 27 November 2025
- Published by Wiley in Scandinavian Journal of Immunology
- Vol. 102 (6) , e70073
The Impact of Complement Factor H‐Related Protein Gene Deletions on Kidney Transplantation
We recently reported that a homozygous deletion in the complement factor H‐related ( CFHR ) locus predisposed kidney transplant patients to rejection. As donors carried intact genes, the susceptibility may have resulted from an alloimmune reaction to FHR proteins. However, we found no evidence for an anti‐FH response. It is therefore possible that CFHR deletions as such affect the rejection risk. Here, we used MLPA and WGS to fine‐map and sequence the CFHR region in rs7542235‐GG patients, a SNP tagging for Δ CFHR3–1 deletion. Our results confirmed that all patients with this SNP harboured deletions of various sizes encompassing CFHR1 . Furthermore, patients with homozygous Δ CFHR3–1 were homozygous for rs6677604‐A, a SNP tagging for deletions of CFHR3–1 locus, confirming that allele A tags for deletion of both CFHR3 and CFHR1 . Proteomics analyses in a larger population demonstrated that rs7542235‐G and rs6677604‐A associate with expression levels of several proteins involved in regulating alloimmune response. We observed that while increasing the rejection risk, the Δ CFHR3–1 did not associate with baseline disease or specific clinical characteristics. To conclude, the various deletion types found in patients shared the deletion of the CFHR1 gene confirming its association with variant rs7542235. Also, both deletion‐tagging alleles are associated with altered expression of FHR proteins.
- 27 November 2025
- Published by Wiley in Scandinavian Journal of Immunology
- Vol. 102 (6) , e70073
Open AccessThe Impact of Complement Factor H‐Related Protein Gene Deletions on Kidney Transplantation
We recently reported that a homozygous deletion in the complement factor H‐related ( CFHR ) locus predisposed kidney transplant patients to rejection. As donors carried intact genes, the susceptibility may have resulted from an alloimmune reaction to FHR proteins. However, we found no evidence for an anti‐FH response. It is therefore possible that CFHR deletions as such affect the rejection risk. Here, we used MLPA and WGS to fine‐map and sequence the CFHR region in rs7542235‐GG patients, a SNP tagging for Δ CFHR3–1 deletion. Our results confirmed that all patients with this SNP harboured deletions of various sizes encompassing CFHR1 . Furthermore, patients with homozygous Δ CFHR3–1 were homozygous for rs6677604‐A, a SNP tagging for deletions of CFHR3–1 locus, confirming that allele A tags for deletion of both CFHR3 and CFHR1 . Proteomics analyses in a larger population demonstrated that rs7542235‐G and rs6677604‐A associate with expression levels of several proteins involved in regulating alloimmune response. We observed that while increasing the rejection risk, the Δ CFHR3–1 did not associate with baseline disease or specific clinical characteristics. To conclude, the various deletion types found in patients shared the deletion of the CFHR1 gene confirming its association with variant rs7542235. Also, both deletion‐tagging alleles are associated with altered expression of FHR proteins.
- Open Access
- 4 September 2025
- Published by Elsevier in Transplantation and Cellular Therapy
- Vol. 31 (12) , 954.e1-954.e4
Impact of MICA-129 Mismatch on Hematopoietic Stem Cell Transplantation Outcomes: Evidence from a Large European Cohort and Meta-Analysis
- 4 September 2025
- Published by Elsevier in Transplantation and Cellular Therapy
- Vol. 31 (12) , 954.e1-954.e4
Open AccessImpact of MICA-129 Mismatch on Hematopoietic Stem Cell Transplantation Outcomes: Evidence from a Large European Cohort and Meta-Analysis
- Open Access
- 15 July 2025
- Published by Wolters Kluwer Health in Transplantation
- Vol. 110 (1) , e175-e184
The Impact of Genome-wide Histocompatibility on Liver Transplantation Outcomes
Background.: Liver transplantation (LT) is the standard treatment for end-stage liver diseases. However, the role of HLA matching in LT remains unclear. Immunological allograft injury and rejection are ongoing concerns, particularly when efforts are made to minimize immunosuppression. Although HLA matching currently has no established role in LT, interest in non-HLA compatibility in the field of transplantation is growing. We compared 666 LT recipient-donor pairs and identified amino acid-changing genetic mismatches outside the HLA gene segment in different protein groups and mismatches in 40 common gene deletions. Methods.: We evaluated the association between mismatches and LT outcomes by using adjusted Cox models for missense variant mismatches and deletion analyses. The primary endpoints were acute rejection, late rejection, graft loss, and overall survival. Statistical significance was set at a false discovery rate (FDR) of P = 0.011; FDR P = 0.001; FDR <0.05). Conclusions.: Genome-wide mismatches outside of the HLA region contribute to the risk of LT complications. Robust, large-scale studies are required to validate these results.
- 15 July 2025
- Published by Wolters Kluwer Health in Transplantation
- Vol. 110 (1) , e175-e184
Open AccessThe Impact of Genome-wide Histocompatibility on Liver Transplantation Outcomes
Background.: Liver transplantation (LT) is the standard treatment for end-stage liver diseases. However, the role of HLA matching in LT remains unclear. Immunological allograft injury and rejection are ongoing concerns, particularly when efforts are made to minimize immunosuppression. Although HLA matching currently has no established role in LT, interest in non-HLA compatibility in the field of transplantation is growing. We compared 666 LT recipient-donor pairs and identified amino acid-changing genetic mismatches outside the HLA gene segment in different protein groups and mismatches in 40 common gene deletions. Methods.: We evaluated the association between mismatches and LT outcomes by using adjusted Cox models for missense variant mismatches and deletion analyses. The primary endpoints were acute rejection, late rejection, graft loss, and overall survival. Statistical significance was set at a false discovery rate (FDR) of P = 0.011; FDR P = 0.001; FDR <0.05). Conclusions.: Genome-wide mismatches outside of the HLA region contribute to the risk of LT complications. Robust, large-scale studies are required to validate these results.
- Open Access
Array Genotyping of Transfusion Relevant Blood Cell Antigens in 6946 Ancestrally Diverse Subjects
Blood transfusions save millions of lives worldwide each year, yet formation of antibodies against non-self antigens remains a significant problem, particularly in frequently transfused patients. We designed and tested the Universal Blood Donor Typing (UBDT_PC1) array for automated high-throughput simultaneous typing of human erythroid, platelet, leukocyte, and neutrophil antigens (HEA, HPA, HLA, and HNA, respectively) to support selection of blood products matched beyond ABO/Rh. Typing samples from 6946 donors of European, African, Admixed American, South Asian, and East Asian ancestry at two different laboratories showed a genotype reproducibility of ≥99% for 17 244 variants, translating to 99.98%, 99.90%, and 99.93% concordance across 338 372 HEA, 53 270 HPA, and 107 094 HLA genotypes, respectively. Compared to previous clinical typing data, concordance was 99.9% and 99.6% for 245 874 HEA and 3726 HPA comparisons, respectively. HLA types were 99.1% concordant with clinical typing across 8130 comparisons, with imputation accuracy higher in Europeans versus non-Europeans. Seven variant RHD alleles, a GYPB deletion underlying the U− phenotype, and 14 high-frequency antigen negative types were also detected. Beyond blood typing, hereditary hemochromatosis-associated HFE variants were identified in 276 donors. We found that the UBDT_PC1 array can reliably type a wide range of blood cell antigens across diverse ancestries. Reproducibility and accuracy were retained when transfusion-relevant targets from the UBDT_PC1 array were incorporated into the UKBB_v2.2 genome-wide typing array. The results represent the potential for significant advancement towards improved patient care by reducing harm in transfused patients through extended matching.
Open AccessArray Genotyping of Transfusion Relevant Blood Cell Antigens in 6946 Ancestrally Diverse Subjects
Blood transfusions save millions of lives worldwide each year, yet formation of antibodies against non-self antigens remains a significant problem, particularly in frequently transfused patients. We designed and tested the Universal Blood Donor Typing (UBDT_PC1) array for automated high-throughput simultaneous typing of human erythroid, platelet, leukocyte, and neutrophil antigens (HEA, HPA, HLA, and HNA, respectively) to support selection of blood products matched beyond ABO/Rh. Typing samples from 6946 donors of European, African, Admixed American, South Asian, and East Asian ancestry at two different laboratories showed a genotype reproducibility of ≥99% for 17 244 variants, translating to 99.98%, 99.90%, and 99.93% concordance across 338 372 HEA, 53 270 HPA, and 107 094 HLA genotypes, respectively. Compared to previous clinical typing data, concordance was 99.9% and 99.6% for 245 874 HEA and 3726 HPA comparisons, respectively. HLA types were 99.1% concordant with clinical typing across 8130 comparisons, with imputation accuracy higher in Europeans versus non-Europeans. Seven variant RHD alleles, a GYPB deletion underlying the U− phenotype, and 14 high-frequency antigen negative types were also detected. Beyond blood typing, hereditary hemochromatosis-associated HFE variants were identified in 276 donors. We found that the UBDT_PC1 array can reliably type a wide range of blood cell antigens across diverse ancestries. Reproducibility and accuracy were retained when transfusion-relevant targets from the UBDT_PC1 array were incorporated into the UKBB_v2.2 genome-wide typing array. The results represent the potential for significant advancement towards improved patient care by reducing harm in transfused patients through extended matching.
- Open Access
- 15 March 2025
- Published by Wiley in Liver International
- Vol. 45 (4) , e70067
Genetic Risk Factors for Steatotic Liver Disease After Liver Transplantation
Background: The role of genetic risk factors for steatotic liver disease (SLD) is intriguing in liver transplantation (LT), as both donor and recipient genetic factors may play a role. There are only a few small‐scale studies published so far. Methods: We analysed the incidence and risk factors for post‐LT SLD and the impact of 56 SLD‐associated genetic variants in 595 donor‐recipient pairs with liver biopsy available ≥ 6 months after LT. We evaluated whether the polygenic risk score (PRS‐5) improves the ability to predict post‐LT SLD in addition to non‐genetic risk factors. Results: SLD after LT was diagnosed in 34.5% of patients during a median 7.6‐year follow‐up. In multivariate analysis including non‐genetic risk factors, donor PNPLA3 rs738409‐G (HR for SLD: C/G 1.34, p = 0.051, G/G 2.25, p = 0.004), donor HSD17B13 rs72613567‐TA (HR for SLD: TA/T 0.68, p = 0.02 TA/TA HR 0.50, p = 0.10) and recipient UCP2 rs695366‐G (HR for SLD: A/G 0.63, p = 0.002, G/G HR 0.50, p = 0.04) appeared as the most important genetic risk factors for post‐LT SLD. The addition of PRS‐5 to a multivariate regression model (including non‐genetic risk factors) improved the predictive ability for SLD only modestly (AUC 0.78 to 0.80). Conclusions: Various genetic variants contribute to post‐LT SLD with separate variants among recipients and donors, with donor PNPLA3 rs738409‐G as the most significant risk allele. Still, donor and recipient genotyping provide only modest additional value for individual risk stratification over phenotype data, highlighting the role of modifiable risk factors.
- 15 March 2025
- Published by Wiley in Liver International
- Vol. 45 (4) , e70067
Open AccessGenetic Risk Factors for Steatotic Liver Disease After Liver Transplantation
Background: The role of genetic risk factors for steatotic liver disease (SLD) is intriguing in liver transplantation (LT), as both donor and recipient genetic factors may play a role. There are only a few small‐scale studies published so far. Methods: We analysed the incidence and risk factors for post‐LT SLD and the impact of 56 SLD‐associated genetic variants in 595 donor‐recipient pairs with liver biopsy available ≥ 6 months after LT. We evaluated whether the polygenic risk score (PRS‐5) improves the ability to predict post‐LT SLD in addition to non‐genetic risk factors. Results: SLD after LT was diagnosed in 34.5% of patients during a median 7.6‐year follow‐up. In multivariate analysis including non‐genetic risk factors, donor PNPLA3 rs738409‐G (HR for SLD: C/G 1.34, p = 0.051, G/G 2.25, p = 0.004), donor HSD17B13 rs72613567‐TA (HR for SLD: TA/T 0.68, p = 0.02 TA/TA HR 0.50, p = 0.10) and recipient UCP2 rs695366‐G (HR for SLD: A/G 0.63, p = 0.002, G/G HR 0.50, p = 0.04) appeared as the most important genetic risk factors for post‐LT SLD. The addition of PRS‐5 to a multivariate regression model (including non‐genetic risk factors) improved the predictive ability for SLD only modestly (AUC 0.78 to 0.80). Conclusions: Various genetic variants contribute to post‐LT SLD with separate variants among recipients and donors, with donor PNPLA3 rs738409‐G as the most significant risk allele. Still, donor and recipient genotyping provide only modest additional value for individual risk stratification over phenotype data, highlighting the role of modifiable risk factors.
- Open Access
The impact of genome-wide histocompatibility on liver transplantation outcomes
Liver transplantation (LT) is the standard treatment for end-stage liver diseases. However, the role of human leukocyte antigen (HLA)-matching in LT remains unclear. Immunological allograft injury and rejection are ongoing concerns, particularly when efforts are made to minimize immunosuppression. Although HLA matching currently has no established role in LT, interest in non-HLA compatibility in the field of transplantation is growing. We compared 666 LT recipient-donor pairs and identified amino acid-changing genetic mismatches outside the HLA gene segment in different protein groups, and mismatches in 40 common gene deletions. We evaluated the association between mismatches and LT outcomes by using adjusted Cox models for missense variant mismatches and deletion analyses. The primary endpoints were acute rejection, late rejection, graft loss, and overall survival. Statistical significance was set at a false discovery rate (FDR) P = 0.011, FDR P = 0.001, FDR <0.05). Genome-wide mismatches outside of the HLA region contribute to the risk of LT complications. Robust, large-scale studies are required to validate these results.
Open AccessThe impact of genome-wide histocompatibility on liver transplantation outcomes
Liver transplantation (LT) is the standard treatment for end-stage liver diseases. However, the role of human leukocyte antigen (HLA)-matching in LT remains unclear. Immunological allograft injury and rejection are ongoing concerns, particularly when efforts are made to minimize immunosuppression. Although HLA matching currently has no established role in LT, interest in non-HLA compatibility in the field of transplantation is growing. We compared 666 LT recipient-donor pairs and identified amino acid-changing genetic mismatches outside the HLA gene segment in different protein groups, and mismatches in 40 common gene deletions. We evaluated the association between mismatches and LT outcomes by using adjusted Cox models for missense variant mismatches and deletion analyses. The primary endpoints were acute rejection, late rejection, graft loss, and overall survival. Statistical significance was set at a false discovery rate (FDR) P = 0.011, FDR P = 0.001, FDR <0.05). Genome-wide mismatches outside of the HLA region contribute to the risk of LT complications. Robust, large-scale studies are required to validate these results.
- Open Access
- 21 March 2024
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 20 (3) , e1011977
A machine-learning method for biobank-scale genetic prediction of blood group antigens
A key element for successful blood transfusion is compatibility of the patient and donor red blood cell (RBC) antigens. Precise antigen matching reduces the risk for immunization and other adverse transfusion outcomes. RBC antigens are encoded by specific genes, which allows developing computational methods for determining antigens from genomic data. We describe here a classification method for determining RBC antigens from genotyping array data. Random forest models for 39 RBC antigens in 14 blood group systems and for human platelet antigen (HPA)-1 were trained and tested using genotype and RBC antigen and HPA-1 typing data available for 1,192 blood donors in the Finnish Blood Service Biobank. The algorithm and models were further evaluated using a validation cohort of 111,667 Danish blood donors. In the Finnish test data set, the median (interquartile range [IQR]) balanced accuracy for 39 models was 99.9 (98.9–100)%. We were able to replicate 34 out of 39 Finnish models in the Danish cohort and the median (IQR) balanced accuracy for classifications was 97.1 (90.1–99.4)%. When applying models trained with the Danish cohort, the median (IQR) balanced accuracy for the 40 Danish models in the Danish test data set was 99.3 (95.1–99.8)%. The RBC antigen and HPA-1 prediction models demonstrated high overall accuracies suitable for probabilistic determination of blood groups and HPA-1 at biobank-scale. Furthermore, population-specific training cohort increased the accuracies of the models. This stand-alone and freely available method is applicable for research and screening for antigen-negative blood donors.
- 21 March 2024
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 20 (3) , e1011977
Open AccessA machine-learning method for biobank-scale genetic prediction of blood group antigens
A key element for successful blood transfusion is compatibility of the patient and donor red blood cell (RBC) antigens. Precise antigen matching reduces the risk for immunization and other adverse transfusion outcomes. RBC antigens are encoded by specific genes, which allows developing computational methods for determining antigens from genomic data. We describe here a classification method for determining RBC antigens from genotyping array data. Random forest models for 39 RBC antigens in 14 blood group systems and for human platelet antigen (HPA)-1 were trained and tested using genotype and RBC antigen and HPA-1 typing data available for 1,192 blood donors in the Finnish Blood Service Biobank. The algorithm and models were further evaluated using a validation cohort of 111,667 Danish blood donors. In the Finnish test data set, the median (interquartile range [IQR]) balanced accuracy for 39 models was 99.9 (98.9–100)%. We were able to replicate 34 out of 39 Finnish models in the Danish cohort and the median (IQR) balanced accuracy for classifications was 97.1 (90.1–99.4)%. When applying models trained with the Danish cohort, the median (IQR) balanced accuracy for the 40 Danish models in the Danish test data set was 99.3 (95.1–99.8)%. The RBC antigen and HPA-1 prediction models demonstrated high overall accuracies suitable for probabilistic determination of blood groups and HPA-1 at biobank-scale. Furthermore, population-specific training cohort increased the accuracies of the models. This stand-alone and freely available method is applicable for research and screening for antigen-negative blood donors.