Tissue Antigens ISSN 0001-2815
LETTER TO THE EDITOR
NO to obsolete definitions: YES to blanks J. M. Nunes, S. Buhler & A. Sanchez-Mazas Laboratory of Anthropology, Genetics and Peopling history (AGP lab), Department of Genetics and Evolution, Anthropology Unit, University of Geneva, Geneva, Switzerland Key words: black; blank; blank allele; Caucasian; data formats; ethnic classification; hispanic; races
Your journal recently published an article by Barbujani (1) clearly showing that it is inappropriate to use ill-defined classifications such as Caucasians, Hispanics or Blacks [see also (2) where this issue is discussed]. These classifications are not only simplistic and incompatible with our knowledge on human diversity, but also overlap, hence invalidating sound and reproducible results. This is why researchers must say NO to obsolete definitions. Unfortunately, these kinds of racial categories are currently used in large scientific networks and databases such as IMGT, NMDP and allelefrequencies.net, with a significant impact in our domain. Such inappropriate practice is common in countries (such as USA and UK) where legal or administrative ethnic classifications exist. Their use is mandatory for government agencies and, by osmosis, for groups or institutions in close relationship with them. The constraint to abide by the legal and administrative racial/ethnic classification is further enforced by a putative financial support. We do not discuss the soundness of integrative policies based on such (almost always self-reported) racial/ethnic classifications, nor their mandatory use in the countries where it is imposed by the law. However, the history of science is full of examples that all go in the same direction and show that a legal reason is not a valid scientific argument. Furthermore, it is not a classification upon which research in population genetics can be based. It would be detrimental to international research that scientists living in countries where such legal definitions do not exist were constrained by their peers to use them. The key point is that the definition of human races is a legacy of colonial anthropology which does not correspond to modern scientific knowledge on human biological variation for which thousands of population samples have been analysed. Since at least half a century and seminal works by renowned geneticists like Luca Cavalli-Sforza, Robert Sokal and their successors, population genetic studies have clearly showed that geography is the best predictor of human genetic variation worldwide, and that this variation follows a continuous pattern. The use of large simplistic human classifications rather than precise populations to define ‘control’ samples in disease-association studies also probably explains why many of these associations fail to be confirmed. If we expect individualised medicine and its promises to become real in a near future, it is mandatory to move from such obsolete categories. © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd Tissue Antigens, 2014, 83, 119–120
Although apparently disconnected, a related methodological issue is the use of a ‘blank’ allele in the description of the HLA genetic profile of a given population. A blank allele is used to represent (under a single name) all alleles that are not observable in a phenotype. The reasons for the occurrence of blank alleles are mostly related to the typing techniques. With most methods, indeed, only a subset of all (known or unknown, often called ‘new’) alleles can be identified (e.g. by using a limited set of DNA probes); hence all the alleles that cannot be identified are grouped within a blank category. A blank allele thus simply occurs when the typings report a single allele instead of two in one or more phenotypes and it is not possible to be completely sure that the corresponding individuals are homozygous. Family studies and/or further typings with more sophisticated techniques, such as cloning-sequencing of full-length genomic DNA and cDNA of HLA genes, may allow to identify ‘true’ homozygous individuals. However, these are rarely applied in current studies. Acknowledging this limitation of common typing techniques should result in the general practice of considering phenotypes exhibiting a single allele as either homozygous or heterozygous with a blank allele. Contrary to what is sometimes believed, the presence of a blank allele has nothing to do with a typing that failed or was not performed. Individuals not typed, if not removed from a sample, introduce much more ambiguous situations, because all possible genotypes for the alleles tested (in addition of a blank!) are then potentially possible. It is also important to remind that the use of blank alleles is a practice that dates back to the early days of HLA data analysis and persisted even when DNA typing was extensively used to type populations [11th and 12th histocompatibility workshops, (3, 4)]. It was unfortunately abandoned in 2002 [except by us, see (5)] for the population data analyses of the 13th histocompatibility workshop. Now let us consider, just to make a small mental experiment, that blank alleles could be dismissed. Let us move back to 2002 and look at frequency estimates made at that time. If these frequencies do not take into account a possible blank allele, then they must be incorrect if new alleles, identified after 2002, have significant frequencies. This is certainly the case of several of the 8068 alleles discovered since IMGT release 1.13 (January 2002 with 1496 alleles identified at that time) and the actual current version of the database (release 119
J. M. Nunes et al.
Letter to the Editor
3.13 of July 2013 with 9564 alleles currently reported, i.e. more than a sixfold increase), although most ‘new’ alleles are actually subtypes of previously identified alleles (third and fourth levels). If the frequencies are estimated using a blank allele, this kind of problem does not arise (or has a much smaller effect) because taking into account the new alleles just leads to a decrease of the frequency of the blank allele. If we now consider the consequences of not using a blank allele in the future, it is easy to predict that either no new alleles discovered after 2013 will have significant frequencies or, alternatively, all past populations frequencies are going to change, possibly a lot, because of the inclusion of new alleles in new versions of the typing kits. In both situations there is an important loss. Clearly, not using a blank is a ‘loss-loss’ situation. To make our argument more poignant, let us consider a case were blank has not been used and, we know it clearly today, should have been. Allele DRB1*14:01 , was first reported in August 1989, and DRB1*14:54 in October 2005. The two alleles are supposed to have been conflated until 2010 as DRB1*14:01 . According to this journal (6), in Germany the frequencies of DRB1*14:54 is nine times more frequent than DRB1*14:01 . Solberg (7) in 2008 does not report DRB1*14:54 (3 years after its official assignment) and attributes a frequency between 2.5% and 7.5% for DRB1*14:01 in Germany. How to reconcile both results? Either the former DRB1*14:01 is reduced ninefold to less than 1% or the sum of allele frequencies for DRB1 must have been significantly different from 1. Both situations are unsatisfactory. Had a blank allele been used, it would have accounted, at least partially, for the frequency of the, yet unidentifiable, DRB1*14:54 , hence making the 2008 frequency estimates of DRB1*14:01 more coherent with those that will be obtained today. Similar stories, where estimated allele frequencies without blank allele are probably no longer reliable, happen with new alleles and ambiguously identified alleles. A further confirmation of the usefulness of using blanks is that most frequency estimations made in the serological or earlier molecular typing times are still largely compatible with current estimates. Then the blank was used, and this is why researchers must say YES to ‘blank’. Concluding; oversimplification is not a good scientific practice. We should use the best science possible, especially when it can be seen that bad practices, be they obsolete ethnicity classifications or the disregard of a potential blank allele, lead to irreproducible results that lack time consistency. For the long-term consistency and coherence of our descriptions of human leukocyte antigen (HLA) (and other genetic systems to which the above considerations may also apply), the use of blanks is required when the homozygosity cannot be granted. Furthermore, simplifying the data to avoid blanks is not only unacceptable but also not necessary because several computer programs exist that allow state-of-the-art analyses of
120
data containing blanks [for instance, the well-known Arlequin (8), or ours Gene[rate] (9)]. Correspondence J.M. Nunes Laboratory of Anthropology Genetics and Peopling history (AGP Lab) Department of Genetics and Evolution Anthropology Unit University of Geneva 12, rue Gustave-Revilliod Geneva Switzerland Tel: +41 223 797 745 Fax: +41 223 793 194 e-mail: [email protected] doi: 10.1111/tan.12276
References 1. Barbujani G, Ghirotto S, Tassi F. Nine things to remember about human genome diversity. Tissue Antigens 2013: 82: 155–64. 2. Sanchez-Mazas A, Vidan-Jeras B, Nunes JM et al. Strategies to work with HLA data in human populations for histocompatibility, clinical transplantation, epidemiology and population genetics: HLA-NET methodological recommendations. Int J Immunogenet 2012: 39: 459–472; quiz 473–476. 3. Imanishi T, Akaza T, Kimura A, Tokunaga K, Gojobori T. Allele and haplotype frequencies for HLA and complement loci in various ethnic groups. In: Tsuji K, Aizawa M, Sasazuki T, eds. HLA 1991 . Oxford : Oxford University Press, 1992 , 1065 –220 . 4. Clayton J, Lonjou C, Whittle D. Allele and haplotype frequencies for HLA loci in various ethnic groups. In: Charron D, ed. Genetic Diversity of HLA Functional and Medical Implications 12th International Histocompatability Workshop Conference Paris June 1996 . Paris: EDK , 1997 , 665 –820 . 5. Sanchez-Mazas A. 13th International Histocompatibility Workshop Anthropology/Human Genetic Diversity Joint Report – Chapter 7: HLA genetic differentiation of the 13th IHWC population data relative to worldwide linguistic families. In: Hansen J, ed. Immunobiology of the Human MHC: Proceedings of the 13th International Histocompatability Workshop and Conference. Seattle, WA: IHWG Press, 2006 , 758 –66 . 6. F¨urst D, Solgi G, Schrezenmeier H, Mytilineos J. The frequency of DRB1*1454 in South German Caucasians. Tissue Antigens 2010: 76: 57–9. 7. Solberg OD, Mack SJ, Lancaster AK et al. Balancing selection and heterogeneity across the classical human leukocyte antigen loci: a meta-analytic review of 497 population studies. Hum Immunol 2008: 69: 443–64. 8. Excoffier L, Lischer HEL. Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Mol Ecol Resour 2010: 10: 564–7. 9. Nunes JM. Gene[rate] [Internet]. Gene[rate]. 2006 Available from: http://geneva.unige.ch/generate/ [accessed September 20, 2013].
© 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd Tissue Antigens, 2014, 83, 119–120
LETTER TO THE EDITOR
NO to obsolete definitions: YES to blanks J. M. Nunes, S. Buhler & A. Sanchez-Mazas Laboratory of Anthropology, Genetics and Peopling history (AGP lab), Department of Genetics and Evolution, Anthropology Unit, University of Geneva, Geneva, Switzerland Key words: black; blank; blank allele; Caucasian; data formats; ethnic classification; hispanic; races
Your journal recently published an article by Barbujani (1) clearly showing that it is inappropriate to use ill-defined classifications such as Caucasians, Hispanics or Blacks [see also (2) where this issue is discussed]. These classifications are not only simplistic and incompatible with our knowledge on human diversity, but also overlap, hence invalidating sound and reproducible results. This is why researchers must say NO to obsolete definitions. Unfortunately, these kinds of racial categories are currently used in large scientific networks and databases such as IMGT, NMDP and allelefrequencies.net, with a significant impact in our domain. Such inappropriate practice is common in countries (such as USA and UK) where legal or administrative ethnic classifications exist. Their use is mandatory for government agencies and, by osmosis, for groups or institutions in close relationship with them. The constraint to abide by the legal and administrative racial/ethnic classification is further enforced by a putative financial support. We do not discuss the soundness of integrative policies based on such (almost always self-reported) racial/ethnic classifications, nor their mandatory use in the countries where it is imposed by the law. However, the history of science is full of examples that all go in the same direction and show that a legal reason is not a valid scientific argument. Furthermore, it is not a classification upon which research in population genetics can be based. It would be detrimental to international research that scientists living in countries where such legal definitions do not exist were constrained by their peers to use them. The key point is that the definition of human races is a legacy of colonial anthropology which does not correspond to modern scientific knowledge on human biological variation for which thousands of population samples have been analysed. Since at least half a century and seminal works by renowned geneticists like Luca Cavalli-Sforza, Robert Sokal and their successors, population genetic studies have clearly showed that geography is the best predictor of human genetic variation worldwide, and that this variation follows a continuous pattern. The use of large simplistic human classifications rather than precise populations to define ‘control’ samples in disease-association studies also probably explains why many of these associations fail to be confirmed. If we expect individualised medicine and its promises to become real in a near future, it is mandatory to move from such obsolete categories. © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd Tissue Antigens, 2014, 83, 119–120
Although apparently disconnected, a related methodological issue is the use of a ‘blank’ allele in the description of the HLA genetic profile of a given population. A blank allele is used to represent (under a single name) all alleles that are not observable in a phenotype. The reasons for the occurrence of blank alleles are mostly related to the typing techniques. With most methods, indeed, only a subset of all (known or unknown, often called ‘new’) alleles can be identified (e.g. by using a limited set of DNA probes); hence all the alleles that cannot be identified are grouped within a blank category. A blank allele thus simply occurs when the typings report a single allele instead of two in one or more phenotypes and it is not possible to be completely sure that the corresponding individuals are homozygous. Family studies and/or further typings with more sophisticated techniques, such as cloning-sequencing of full-length genomic DNA and cDNA of HLA genes, may allow to identify ‘true’ homozygous individuals. However, these are rarely applied in current studies. Acknowledging this limitation of common typing techniques should result in the general practice of considering phenotypes exhibiting a single allele as either homozygous or heterozygous with a blank allele. Contrary to what is sometimes believed, the presence of a blank allele has nothing to do with a typing that failed or was not performed. Individuals not typed, if not removed from a sample, introduce much more ambiguous situations, because all possible genotypes for the alleles tested (in addition of a blank!) are then potentially possible. It is also important to remind that the use of blank alleles is a practice that dates back to the early days of HLA data analysis and persisted even when DNA typing was extensively used to type populations [11th and 12th histocompatibility workshops, (3, 4)]. It was unfortunately abandoned in 2002 [except by us, see (5)] for the population data analyses of the 13th histocompatibility workshop. Now let us consider, just to make a small mental experiment, that blank alleles could be dismissed. Let us move back to 2002 and look at frequency estimates made at that time. If these frequencies do not take into account a possible blank allele, then they must be incorrect if new alleles, identified after 2002, have significant frequencies. This is certainly the case of several of the 8068 alleles discovered since IMGT release 1.13 (January 2002 with 1496 alleles identified at that time) and the actual current version of the database (release 119
J. M. Nunes et al.
Letter to the Editor
3.13 of July 2013 with 9564 alleles currently reported, i.e. more than a sixfold increase), although most ‘new’ alleles are actually subtypes of previously identified alleles (third and fourth levels). If the frequencies are estimated using a blank allele, this kind of problem does not arise (or has a much smaller effect) because taking into account the new alleles just leads to a decrease of the frequency of the blank allele. If we now consider the consequences of not using a blank allele in the future, it is easy to predict that either no new alleles discovered after 2013 will have significant frequencies or, alternatively, all past populations frequencies are going to change, possibly a lot, because of the inclusion of new alleles in new versions of the typing kits. In both situations there is an important loss. Clearly, not using a blank is a ‘loss-loss’ situation. To make our argument more poignant, let us consider a case were blank has not been used and, we know it clearly today, should have been. Allele DRB1*14:01 , was first reported in August 1989, and DRB1*14:54 in October 2005. The two alleles are supposed to have been conflated until 2010 as DRB1*14:01 . According to this journal (6), in Germany the frequencies of DRB1*14:54 is nine times more frequent than DRB1*14:01 . Solberg (7) in 2008 does not report DRB1*14:54 (3 years after its official assignment) and attributes a frequency between 2.5% and 7.5% for DRB1*14:01 in Germany. How to reconcile both results? Either the former DRB1*14:01 is reduced ninefold to less than 1% or the sum of allele frequencies for DRB1 must have been significantly different from 1. Both situations are unsatisfactory. Had a blank allele been used, it would have accounted, at least partially, for the frequency of the, yet unidentifiable, DRB1*14:54 , hence making the 2008 frequency estimates of DRB1*14:01 more coherent with those that will be obtained today. Similar stories, where estimated allele frequencies without blank allele are probably no longer reliable, happen with new alleles and ambiguously identified alleles. A further confirmation of the usefulness of using blanks is that most frequency estimations made in the serological or earlier molecular typing times are still largely compatible with current estimates. Then the blank was used, and this is why researchers must say YES to ‘blank’. Concluding; oversimplification is not a good scientific practice. We should use the best science possible, especially when it can be seen that bad practices, be they obsolete ethnicity classifications or the disregard of a potential blank allele, lead to irreproducible results that lack time consistency. For the long-term consistency and coherence of our descriptions of human leukocyte antigen (HLA) (and other genetic systems to which the above considerations may also apply), the use of blanks is required when the homozygosity cannot be granted. Furthermore, simplifying the data to avoid blanks is not only unacceptable but also not necessary because several computer programs exist that allow state-of-the-art analyses of
120
data containing blanks [for instance, the well-known Arlequin (8), or ours Gene[rate] (9)]. Correspondence J.M. Nunes Laboratory of Anthropology Genetics and Peopling history (AGP Lab) Department of Genetics and Evolution Anthropology Unit University of Geneva 12, rue Gustave-Revilliod Geneva Switzerland Tel: +41 223 797 745 Fax: +41 223 793 194 e-mail: [email protected] doi: 10.1111/tan.12276
References 1. Barbujani G, Ghirotto S, Tassi F. Nine things to remember about human genome diversity. Tissue Antigens 2013: 82: 155–64. 2. Sanchez-Mazas A, Vidan-Jeras B, Nunes JM et al. Strategies to work with HLA data in human populations for histocompatibility, clinical transplantation, epidemiology and population genetics: HLA-NET methodological recommendations. Int J Immunogenet 2012: 39: 459–472; quiz 473–476. 3. Imanishi T, Akaza T, Kimura A, Tokunaga K, Gojobori T. Allele and haplotype frequencies for HLA and complement loci in various ethnic groups. In: Tsuji K, Aizawa M, Sasazuki T, eds. HLA 1991 . Oxford : Oxford University Press, 1992 , 1065 –220 . 4. Clayton J, Lonjou C, Whittle D. Allele and haplotype frequencies for HLA loci in various ethnic groups. In: Charron D, ed. Genetic Diversity of HLA Functional and Medical Implications 12th International Histocompatability Workshop Conference Paris June 1996 . Paris: EDK , 1997 , 665 –820 . 5. Sanchez-Mazas A. 13th International Histocompatibility Workshop Anthropology/Human Genetic Diversity Joint Report – Chapter 7: HLA genetic differentiation of the 13th IHWC population data relative to worldwide linguistic families. In: Hansen J, ed. Immunobiology of the Human MHC: Proceedings of the 13th International Histocompatability Workshop and Conference. Seattle, WA: IHWG Press, 2006 , 758 –66 . 6. F¨urst D, Solgi G, Schrezenmeier H, Mytilineos J. The frequency of DRB1*1454 in South German Caucasians. Tissue Antigens 2010: 76: 57–9. 7. Solberg OD, Mack SJ, Lancaster AK et al. Balancing selection and heterogeneity across the classical human leukocyte antigen loci: a meta-analytic review of 497 population studies. Hum Immunol 2008: 69: 443–64. 8. Excoffier L, Lischer HEL. Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Mol Ecol Resour 2010: 10: 564–7. 9. Nunes JM. Gene[rate] [Internet]. Gene[rate]. 2006 Available from: http://geneva.unige.ch/generate/ [accessed September 20, 2013].
© 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd Tissue Antigens, 2014, 83, 119–120
