Errors in genomics and proteomics
1
CORRESPONDENCE http://biotech.nature.com • JANUARY 2002 • VOLUME 20 • nature biotechnology Transgenes in Mexican maize To the editor: Genetic flow between transgenic and native maize has apparently occurred in Mexico 1 , resulting in wild strains contain- ing one or more transgenic sequences (most likely encoding Bt toxin). These “transgenic” native maizes not only have every single trait that has been selected and preserved for thousands of years (making them perfectly adapted to specific geo- graphic regions), but now also possess an additional and desirable characteristic— insect resistance, a trait likely to be con- sciously preferred by Mexican peasant farmers. Diversity will not be affected. On the contrary, we can predict that this useful transgene will be found in increasing num- bers and types of native maizes. We believe it is important to stress this is not genetic contamination! Contamination means unexpected, undesirable, and uncontrollable spread; that is not happen- ing. The spread will be induced because of the advantage of having a native corn with resistance to insects. Maize is so dependent on human inter- vention that it cannot survive in the wild. Maize seeds are attached to a cob and can- not free themselves: it absolutely requires human intervention. As maize was first domesticated more than 6,000 years ago, only genes and alleles that are important for humans have been selected and pre- served. Still, if someone wants to remove the transgene from these plants, the procedure would be simple: select and multiply those susceptible maizes and do not harvest and multiply the insect-resistant ones. That is something no Mexican farmer will do. Teosintes, ancestors and close relatives of corn, do not seem to be affected by genetic flow from (any) maize. Teosintes growing naturally in cornfields yield a very poor hybrid progeny. They do not release their seeds, and therefore the probability is very low for natural genetic introgression (incorporation of a gene or allele in a pop- ulation) into teosintes. We also have found that teosintes are highly susceptible to insects and pathogens when growing under more intensive experimental field condi- tions, but they appear to be resistant to them when growing naturally in the wild. Thus we conclude that even if the Bt transgene could be introgressed into teosintes, it will provide no biological advantage and thus would be lost by natur- al evolution. To reiterate 2 , there is no need for concern. Juan Pablo Ricardo Martínez-Soriano and Ana María Bailey, Unidad de Biotecnología e Ingeniería Genética de Plantas, Centro de Investigación y Estudios Avanzados del Instituto Politécnico Nacional, Apartado Postal 629, 36500 Irapuato, Gto, México ([email protected]) Joel Lara-Reyna,Instituto de Fitosanidad,Colegio de Postgraduados,Montecillo, Edo de México, México, and Diana Sara Leal-Klevezas Secretaría de Salud de Guanajuato, Tamazuca 4 Centro, Guanajuato, Gto., México 1. Dalton, R. Nature 413, 337 (2001). 2. Martínez-Soriano J.P. & Leal-Klevezas, D.S. Science 287, 1399 (2000). Errors in genomics and proteomics To the editor: Large-scale studies of gene expression on the RNA, protein, and/or metabolite level should greatly help to understand complex biological processes. However, it now becomes apparent that the correlation between mRNA and protein levels is remarkably and unexpectedly low; for example, the moderate correlation between levels of transcript and protein in yeast 1 has confirmed an earlier study using serial analysis of gene expression (SAGE) 2 . This low correlation is generally hypoth- esized to result from post-translational modifications, which therefore seem more frequent than previously assumed. It would also imply that RNA studies are less predic- tive, notably for complex traits, than pro- tein studies are. But should we discount RNA expression profiling and favor proteomics? We think it is too early to con- clude and believe that different distribu- tions of experimental noise can contribute to the lack of correlation between RNA and protein data. In large populations of transformed tobacco plants carrying the β-glu- curonidase (GUS) and the luciferase (LUC) reporter genes, enzyme activities correlate poorly with the respective mRNA levels (R = 0.46–0.67). In contrast, GUS enzyme activities correlate well with LUC activities (R = 0.80), and GUS mRNA levels correlate well with LUC mRNA levels (R = 0.94). As neither of the proteins undergoes any post- translational modification, the counterin- tuitive result of the poor RNA/protein correlation cannot be due to changes at this level. In this case, we postulate the existence of “error pipelines” that mask a biologically relevant correlation. Assays of mRNA, by whatever method, involve experimental steps in common for each mRNA. Each step will have its own errors associated with it. Some of these errors will be system- atic in nature and point in the same direction. The crux is that in correlation analysis, such errors will cancel out, result- ing in better correlations. The same holds for protein assays. However, when mRNA assays are correlated with protein assays, different methods of analysis are combined and the errors do not cancel out. This way, relevant correlations can become blurred by what is essentially experimental noise and not necessarily post-translational modification(s). Our concept of error pipelines predicts that similarly low correlations will be found between genomics or proteomics and future metabolomics data (as in ref. 3). Preventing a blur due to error pipelines requires the careful assessment of the hier- archy in, and quality of, data sets. Reducing experimental noise by further technical improvements and/or increased replica- tion, for example, with the use of segregat- ing populations 4 , will prove important for any associative study of different data types in genomics to become meaningful. Many more mRNA assays may correlate much more nicely with their corresponding pro- tein assays, and the still easier/cheaper large-scale RNA analyses can remain the method of choice. Ritsert C. Jansen and Jan-Peter Nap, Plant Research International, Wageningen University and Research Centre, Wageningen, The Netherlands, and Ludmila Mlynárová Institute of Plant Genetics and Biotechnology, Nitra, Slovak Republic ([email protected]) 1. Ideker, T. et al. Science 292, 929–934 (2001). 2.Gygi, S.P., Rochon, Y., Franza, B.R., & Aebersold, R. Mol. Cell. Biol. 19, 1720–1730 (1999). 3. Fiehn, O. et al. Nat. Biotechnol. 18, 1157–1161 (2000). 4. Jansen, R.C. & Nap, J.P. Trends Genet. 17, 388–391 (2001). 19 Letters may be edited for space and clarity. They should be addressed to: Correspondence Nature Biotechnology 345 Park Avenue South New York, NY 10010-1707, USA or sent by e-mail to [email protected] Please include your telephone and fax numbers. © 2002 Nature Publishing Group http://biotech.nature.com
Transcript of Errors in genomics and proteomics
CORRESPONDENCE
http://biotech.nature.com • JANUARY 2002 • VOLUME 20 • nature biotechnology
Transgenes in Mexican maize
To the editor:Genetic flow between transgenic andnative maize has apparently occurred inMexico1, resulting in wild strains contain-ing one or more transgenic sequences(most likely encoding Bt toxin). These“transgenic” native maizes not only haveevery single trait that has been selected andpreserved for thousands of years (makingthem perfectly adapted to specific geo-graphic regions), but now also possess anadditional and desirable characteristic—insect resistance, a trait likely to be con-sciously preferred by Mexican peasantfarmers. Diversity will not be affected. Onthe contrary, we can predict that this usefultransgene will be found in increasing num-bers and types of native maizes.
We believe it is important to stress this isnot genetic contamination! Contaminationmeans unexpected, undesirable, anduncontrollable spread; that is not happen-ing. The spread will be induced because ofthe advantage of having a native corn withresistance to insects.
Maize is so dependent on human inter-vention that it cannot survive in the wild.Maize seeds are attached to a cob and can-not free themselves: it absolutely requireshuman intervention. As maize was firstdomesticated more than 6,000 years ago,only genes and alleles that are importantfor humans have been selected and pre-served.
Still, if someone wants to remove thetransgene from these plants, the procedurewould be simple: select and multiply thosesusceptible maizes and do not harvest andmultiply the insect-resistant ones. That issomething no Mexican farmer will do.
Teosintes, ancestors and close relatives ofcorn, do not seem to be affected by geneticflow from (any) maize. Teosintes growingnaturally in cornfields yield a very poorhybrid progeny. They do not release theirseeds, and therefore the probability is verylow for natural genetic introgression(incorporation of a gene or allele in a pop-ulation) into teosintes. We also have foundthat teosintes are highly susceptible toinsects and pathogens when growing under
more intensive experimental field condi-tions, but they appear to be resistant tothem when growing naturally in the wild.
Thus we conclude that even if the Bttransgene could be introgressed intoteosintes, it will provide no biologicaladvantage and thus would be lost by natur-al evolution. To reiterate2, there is no needfor concern.
Juan Pablo Ricardo Martínez-Soriano andAna María Bailey,
Unidad de Biotecnología e IngenieríaGenética de Plantas,
Centro de Investigación y Estudios Avanzadosdel Instituto Politécnico Nacional,
Apartado Postal 629, 36500 Irapuato,Gto, México
([email protected])Joel Lara-Reyna,Instituto de
Fitosanidad,Colegio dePostgraduados,Montecillo,
Edo de México, México,and Diana Sara Leal-Klevezas
Secretaría de Salud de Guanajuato,Tamazuca 4 Centro,
Guanajuato,Gto., México
1. Dalton, R. Nature 413, 337 (2001).2. Martínez-Soriano J.P. & Leal-Klevezas, D.S.
Science 287, 1399 (2000).
Errors in genomics and proteomics
To the editor:Large-scale studies of gene expression onthe RNA, protein, and/or metabolite levelshould greatly help to understand complexbiological processes. However, it nowbecomes apparent that the correlationbetween mRNA and protein levels isremarkably and unexpectedly low; forexample, the moderate correlation betweenlevels of transcript and protein in yeast1 hasconfirmed an earlier study using serialanalysis of gene expression (SAGE)2.
This low correlation is generally hypoth-esized to result from post-translationalmodifications, which therefore seem morefrequent than previously assumed. It wouldalso imply that RNA studies are less predic-tive, notably for complex traits, than pro-tein studies are. But should we discountRNA expression profiling and favor proteomics? We think it is too early to con-clude and believe that different distribu-tions of experimental noise can contributeto the lack of correlation between RNA andprotein data.
In large populations of transformedtobacco plants carrying the β-glu-curonidase (GUS) and the luciferase (LUC)reporter genes, enzyme activities correlatepoorly with the respective mRNA levels
(R = 0.46–0.67). In contrast, GUS enzymeactivities correlate well with LUC activities(R = 0.80), and GUS mRNA levels correlatewell with LUC mRNA levels (R = 0.94). Asneither of the proteins undergoes any post-translational modification, the counterin-tuitive result of the poor RNA/protein correlation cannot be due to changes at thislevel.
In this case, we postulate the existence of“error pipelines” that mask a biologicallyrelevant correlation. Assays of mRNA, bywhatever method, involve experimentalsteps in common for each mRNA. Eachstep will have its own errors associatedwith it. Some of these errors will be system-atic in nature and point in the same direction. The crux is that in correlationanalysis, such errors will cancel out, result-ing in better correlations. The same holdsfor protein assays. However, when mRNAassays are correlated with protein assays,different methods of analysis are combinedand the errors do not cancel out. This way,relevant correlations can become blurredby what is essentially experimental noiseand not necessarily post-translationalmodification(s).
Our concept of error pipelines predictsthat similarly low correlations will befound between genomics or proteomicsand future metabolomics data (as in ref. 3).Preventing a blur due to error pipelinesrequires the careful assessment of the hier-archy in, and quality of, data sets. Reducingexperimental noise by further technicalimprovements and/or increased replica-tion, for example, with the use of segregat-ing populations4, will prove important forany associative study of different data typesin genomics to become meaningful. Manymore mRNA assays may correlate muchmore nicely with their corresponding pro-tein assays, and the still easier/cheaperlarge-scale RNA analyses can remain themethod of choice.
Ritsert C. Jansen and Jan-Peter Nap,Plant Research International,
Wageningen University and Research Centre,Wageningen,
The Netherlands,and
Ludmila MlynárováInstitute of Plant Genetics and Biotechnology,
Nitra,Slovak Republic
1. Ideker, T. et al. Science 292, 929–934 (2001).2. Gygi, S.P., Rochon, Y., Franza, B.R., & Aebersold,
R. Mol. Cell. Biol. 19, 1720–1730 (1999).3. Fiehn, O. et al. Nat. Biotechnol. 18, 1157–1161
(2000).4. Jansen, R.C. & Nap, J.P. Trends Genet. 17, 388–391
(2001).
19
Letters may be edited for space and clarity.They should be addressed to:CorrespondenceNature Biotechnology345 Park Avenue SouthNew York, NY 10010-1707, USAor sent by e-mail to [email protected] include your telephone and fax numbers.
©20
02 N
atu
re P
ub
lish
ing
Gro
up
h
ttp
://b
iote
ch.n
atu
re.c
om
