A deep learning approach to prediction of blood group antigens from genomic data.

BACKGROUND: Deep learning methods are revolutionizing natural science. In this study, we aim to apply such techniques to develop blood type prediction models based on cheap to analyze and easily scalable screening array genotyping platforms. METHODS: Combining existing blood types from blood banks and imputed screening array genotypes for ~111,000 Danish and 1168 Finnish blood donors, we used deep learning techniques to train and validate blood type prediction models for 36 antigens in 15 blood group systems. To account for missing genotypes a denoising autoencoder initial step was utilized, followed by a convolutional neural network blood type classifier. RESULTS: Two thirds of the trained blood type prediction models demonstrated an F1-accuracy above 99%. Models for antigens with low or high frequencies like, for example, Cw, low training cohorts like, for example, Cob, or very complicated genetic underpinning like, for example, RhD, proved to be more challenging for high accuracy (>99%) DL modeling. However, in the Danish cohort only 4 out of 36 models (Cob, Cw, D-weak, Kpa) failed to achieve a prediction F1-accuracy above 97%. This high predictive performance was replicated in the Finnish cohort. DISCUSSION: High accuracy in a variety of blood groups proves viability of deep learning-based blood type prediction using array chip genotypes, even in blood groups with nontrivial genetic underpinnings. These techniques are suitable for aiding in identifying blood donors with rare blood types by greatly narrowing down the potential pool of candidate donors before clinical grade confirmation.

SMIM1 absence is associated with reduced energy expenditure and excess weight.

BACKGROUND: Obesity rates have nearly tripled in the past 50 years, and by 2030 more than 1 billion individuals worldwide are projected to be obese. This creates a significant economic strain due to the associated non-communicable diseases. The root cause is an energy expenditure imbalance, owing to an interplay of lifestyle, environmental, and genetic factors. Obesity has a polygenic genetic architecture; however, single genetic variants with large effect size are etiological in a minority of cases. These variants allowed the discovery of novel genes and biology relevant to weight regulation and ultimately led to the development of novel specific treatments. METHODS: We used a case-control approach to determine metabolic differences between individuals homozygous for a loss-of-function genetic variant in the small integral membrane protein 1 (SMIM1) and the general population, leveraging data from five cohorts. Metabolic characterization of SMIM1-/- individuals was performed using plasma biochemistry, calorimetric chamber, and DXA scan. FINDINGS: We found that individuals homozygous for a loss-of-function genetic variant in SMIM1 gene, underlying the blood group Vel, display excess body weight, dyslipidemia, altered leptin to adiponectin ratio, increased liver enzymes, and lower thyroid hormone levels. This was accompanied by a reduction in resting energy expenditure. CONCLUSION: This research identified a novel genetic predisposition to being overweight or obese. It highlights the need to investigate the genetic causes of obesity to select the most appropriate treatment given the large cost disparity between them. FUNDING: This work was funded by the National Institute of Health Research, British Heart Foundation, and NHS Blood and Transplant.

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.

Genetic prediction of 33 blood group phenotypes using an existing genotype dataset.

BACKGROUND: Accurate blood type data are essential for blood bank management, but due to costs, few of 43 blood group systems are routinely determined in Danish blood banks. However, a more comprehensive dataset of blood types is useful in scenarios such as rare blood type allocation. We aimed to investigate the viability and accuracy of predicting blood types by leveraging an existing dataset of imputed genotypes for two cohorts of approximately 90,000 each (Danish Blood Donor Study and Copenhagen Biobank) and present a more comprehensive overview of blood types for our Danish donor cohort. STUDY DESIGN AND METHODS: Blood types were predicted from genome array data using known variant determinants. Prediction accuracy was confirmed by comparing with preexisting serological blood types. The Vel blood group was used to test the viability of using genetic prediction to narrow down the list of candidate donors with rare blood types. RESULTS: Predicted phenotypes showed a high balanced accuracy >99.5% in most cases: A, B, C/c, Coa /Cob , Doa /Dob , E/e, Jka /Jkb , Kna /Knb , Kpa /Kpb , M/N, S/s, Sda , Se, and Yta /Ytb , while some performed slightly worse: Fya /Fyb , K/k, Lua /Lub , and Vel ~99%-98% and CW and P1 ~96%. Genetic prediction identified 70 potential Vel negatives in our cohort, 64 of whom were confirmed correct using polymerase chain reaction (negative predictive value: 91.5%). DISCUSSION: High genetic prediction accuracy in most blood groups demonstrated the viability of generating blood types using preexisting genotype data at no cost and successfully narrowed the pool of potential individuals with the rare Vel-negative phenotype from 180,000 to 70.

A large cohort study of the effects of Lewis, ABO, 13 other blood groups, and secretor status on COVID-19 susceptibility, severity, and long COVID-19.

BACKGROUND: Previous studies have reported Blood type O to confer a lower risk of SARS-CoV-2 infection, while secretor status and other blood groups have been suspected to have a similar effect as well. STUDY DESIGN AND METHODS: To determine whether any other blood groups influence testing positive for SARS-CoV-2, COVID-19 severity, or prolonged COVID-19, we used a large cohort of 650,156 Danish blood donors with varying available data for secretor status and blood groups ABO, Rh, Colton, Duffy, Diego, Dombrock, Kell, Kidd, Knops, Lewis, Lutheran, MNS, P1PK, Vel, and Yt. Of these, 36,068 tested positive for SARS-CoV-2 whereas 614,088 tested negative between 2020-02-17 and 2021-08-04. Associations between infection and blood groups were assessed using logistic regression models with sex and age as covariates. RESULTS: The Lewis blood group antigen Lea displayed strongly reduced SARS-CoV-2 susceptibility OR 0.85 CI[0.79-0.93] p < .001. Compared to blood type O, the blood types B, A, and AB were found more susceptible toward infection with ORs 1.1 CI[1.06-1.14] p < .001, 1.17 CI[1.14-1.2] p < .001, and 1.2 CI[1.14-1.26] p < .001, respectively. No susceptibility associations were found for the other 13 blood groups investigated. There was no association between any blood groups and COVID-19 hospitalization or long COVID-19. No secretor status associations were found. DISCUSSION: This study uncovers a new association to reduced SARS-CoV-2 susceptibility for Lewis type Lea and confirms the previous link to blood group O. The new association to Lea could be explained by a link between mucosal microbiome and SARS-CoV-2.

Symbiosis genes show a unique pattern of introgression and selection within a Rhizobium leguminosarum species complex.

Rhizobia supply legumes with fixed nitrogen using a set of symbiosis genes. These can cross rhizobium species boundaries, but it is unclear how many other genes show similar mobility. Here, we investigate inter-species introgression using de novo assembly of 196 Rhizobium leguminosarum sv. trifolii genomes. The 196 strains constituted a five-species complex, and we calculated introgression scores based on gene-tree traversal to identify 171 genes that frequently cross species boundaries. Rather than relying on the gene order of a single reference strain, we clustered the introgressing genes into four blocks based on population structure-corrected linkage disequilibrium patterns. The two largest blocks comprised 125 genes and included the symbiosis genes, a smaller block contained 43 mainly chromosomal genes, and the last block consisted of three genes with variable genomic location. All introgression events were likely mediated by conjugation, but only the genes in the symbiosis linkage blocks displayed overrepresentation of distinct, high-frequency haplotypes. The three genes in the last block were core genes essential for symbiosis that had, in some cases, been mobilized on symbiosis plasmids. Inter-species introgression is thus not limited to symbiosis genes and plasmids, but other cases are infrequent and show distinct selection signatures.