In transplantation of hematopoietic stem cells (HSCs) from unrelated donors a high HLA compatibility level decreases the risk of acute graft-versus-host disease and mortality. The diversity of the HLA system at the allelic and haplotypic level and the heterogeneity of HLA typing data of the registered donors render the search process a complex task. This paper summarizes our experience with a search algorithm that includes at the start of the search a probability estimate (high/intermediate/low) to identify a HLA-A, B, C, DRB1, DQB1-compatible donor (a 10/10 match). Based on 2002-2011 searches about 30% of patients have a high, 30% an intermediate, and 40% a low probability search. Search success rate and duration are presented and discussed in light of the experience of other centers. Overall a 9-10/10 matched HSC donor can now be identified for 60-80% of patients of European descent. For high probability searches donors can be selected on the basis of DPB1-matching with an estimated success rate of >40%. For low probability searches there is no consensus on which HLA incompatibilities are more permissive, although HLA-DQB1 mismatches are generally considered as acceptable. Models for the discrimination of more detrimental mismatches based on specific amino acid residues rather than specific HLA alleles are presented.
Keyphrases
- end stage renal disease
- chronic kidney disease
- machine learning
- type diabetes
- deep learning
- cardiovascular disease
- hematopoietic stem cell
- prognostic factors
- amino acid
- peritoneal dialysis
- big data
- intensive care unit
- coronary artery disease
- signaling pathway
- patient reported outcomes
- oxidative stress
- risk factors
- electronic health record
- endoplasmic reticulum stress
- cell cycle arrest
- acute respiratory distress syndrome
- patient reported