Basics of QTL analysis - University of Arkansas · Basics of QTL analysis RiceCAP MAS workshop ... Dept. of Plant Pathology, Kansas State University Plan zAim of this presentation

1

Basics of QTL analysisBasics of QTL analysis

RiceCAP MAS workshopJune 14 14 -- 16,16, 20052005

Clare NelsonDept. of Plant Pathology, Kansas State University

PlanPlanAim of this presentationAim of this presentationDescription of the QTL mapping problemDescription of the QTL mapping problemStatistical approachesStatistical approaches

singlesingle--marker analysesmarker analysesinterval analysesinterval analyses

regression/likelihoodregression/likelihoodBayesianBayesian

Software tools (absurdly brief intro)Software tools (absurdly brief intro)Data explorationData exploration

with horror stories, at no extra chargewith horror stories, at no extra charge

2

Aim of this presentationAim of this presentation

Give you a grasp of the statistical basis for Give you a grasp of the statistical basis for QTL analysisQTL analysisGive guidelines for doing the analysis Give guidelines for doing the analysis meaningfullymeaningfully

(without fooling anyone)(without fooling anyone)

AssumptionsAssumptionsYouYou’’re familiar withre familiar with

MendelianMendelian geneticsgeneticslocuslocus, , alleleallele, , phenotypephenotype, , segregation, segregation, homozygous/heterozygous, inbreedinghomozygous/heterozygous, inbreeding

molecular geneticsmolecular geneticsDNADNA--DNA hybridization, DNA DNA hybridization, DNA markermarker, genetic , genetic mapmap, , recombinationrecombination

statisticsstatisticscontrast of two contrast of two meansmeansregression, ANOVAregression, ANOVA

3

Classical (preClassical (pre--molecular) geneticsmolecular) geneticsMotivationMotivation

Explain and cure genetic diseases ("inborn errors of metabolism"Explain and cure genetic diseases ("inborn errors of metabolism"))ApproachApproach

Explain how genotype causes phenotype (so we can avoid or repairExplain how genotype causes phenotype (so we can avoid or repair))MethodMethod

Identify or create a mutant phenotypeIdentify or create a mutant phenotypeIdentify the biochemical pathway and protein containing the lesiIdentify the biochemical pathway and protein containing the lesiononDo geneticsDo genetics

Cross mutant with wild type, observe segregation of trait in tesCross mutant with wild type, observe segregation of trait in testcross or Ftcross or F22progenyprogenyCross mutant with other mutants to establish allelism, effect ofCross mutant with other mutants to establish allelism, effect of two two mutations, epistasis, etc.mutations, epistasis, etc.Perform linkage analysis to place a gene on the genetic Perform linkage analysis to place a gene on the genetic mapmap

DisadvantagesDisadvantagesworks only with simple traits (controlled by 1 or 2 genes)works only with simple traits (controlled by 1 or 2 genes)doesn't yield gene sequence (basic explanation of lesion)doesn't yield gene sequence (basic explanation of lesion)is lowis low--throughput (one gene: one postdoc)throughput (one gene: one postdoc)gives low resolution on mapgives low resolution on map

Classical quantitative geneticsClassical quantitative geneticsMotivationMotivation

Improve crops and livestock by crossing and selectionImprove crops and livestock by crossing and selectionApproachApproach

Predict the means & variances of Predict the means & variances of polygenicpolygenic traitstraitsMethodMethod

Make single, reciprocal, and complex crosses, evaluate means & Make single, reciprocal, and complex crosses, evaluate means & variances in progenyvariances in progenyBased on theory of shared alleles among relatives, determine Based on theory of shared alleles among relatives, determine heritabilityheritabilityof phenotypic traitof phenotypic traitUsing predicted response to selection, adopt suitable breeding dUsing predicted response to selection, adopt suitable breeding designsesigns

AdvantagesAdvantagesGives great practical success for certain kinds of breedingGives great practical success for certain kinds of breedingProvides essential theory underlying molecular quantitative geneProvides essential theory underlying molecular quantitative geneticstics

DisadvantagesDisadvantagesMany traits aren't truly Many traits aren't truly polygenicpolygenic (controlled by large numbers of genes (controlled by large numbers of genes with infinitesimal effect)with infinitesimal effect)Method can't identify individual genes/alleles or explain gene aMethod can't identify individual genes/alleles or explain gene actionction

4

Molecular quantitative geneticsMolecular quantitative geneticsMotivationMotivation

Explain genetic control of complex (multiExplain genetic control of complex (multi--gene) traitsgene) traitsApproachApproach

Obtain genotypes of genetic markersObtain genotypes of genetic markersCompute statistical correlation with phenotypeCompute statistical correlation with phenotype

MethodMethodCross two parents divergent for traits of interestCross two parents divergent for traits of interestAssay phenotype and marker genotypes in BC or FAssay phenotype and marker genotypes in BC or F22 progenyprogenyInfer QTLs from markers showing statistical association with traInfer QTLs from markers showing statistical association with traitit

AdvantagesAdvantagesAccommodates continuously distributed, multigenic traitsAccommodates continuously distributed, multigenic traitsProvides aids for genetic selection (mostly in crops/livestock)Provides aids for genetic selection (mostly in crops/livestock)

DisadvantagesDisadvantagesMarker is not the geneMarker is not the geneIdentifying the true genetic model is still a hard problemIdentifying the true genetic model is still a hard problemPhenotype is the end product of many genetic processes over timePhenotype is the end product of many genetic processes over time

Molecular quantitative geneticsMolecular quantitative genetics

phenotypic variationgenetic (allelic)variation

AA aa

For each DNA marker:For each DNA marker:we contrast mean phenotype of one genotype class vs. otherswe contrast mean phenotype of one genotype class vs. others

AA

aa

AA + aa

5

Terminology quizTerminology quiz

ABCDEFGABCDEFG

abcdefgabcdefg

ABCDEFGabcdefg

ZYXWVUTZYXWVUT

ZYXWVUTzyxwvut

zyxwvutzyxwvut

ABCdeFGZyxwVuT

XX

X

What part of this figure representsWhat part of this figure representslocus?locus?allele?allele?genotype?genotype?

Note: complex crosses like this are harder to analyze Note: complex crosses like this are harder to analyze statistically than simple bistatistically than simple bi--allelic ones.allelic ones.

DNA marker genotyping: RFLPDNA marker genotyping: RFLP

source of autoradiogram image: FujiFilm

P1 P2 F1 1 2 Backcross progeny 9 10 11 12

600 bp

350 bp

Genotype scores of progeny are recordedas # of P1 alleles: 1, 2, 2, 1, 1 for progeny 1 to 5or # of P2 alleles: 1, 0, 0, 1, 1or with any symbol you like: ABBAA for example

6

A marker datasetA marker dataset*ABG704 *ABG704

BBBBBBAABBAAAABABABABABAABAAABAAAABBBBBBBAAAABBBBABBAAAABBBBBBBAABBAAAABABABABABAABAAABAAAABBBBBBBAAAABBBBABBAAAABABBABBABBBBBBBBBAAABABAABBABAAABBAABBABAAAABAABAABBBAABBBABBABBABBBBBBBBBAAABABAABBABAAABBAABBABAAAABAABAABBBAABBBAABBAABBBAABAABBABBBBBBAABAABAABBABBAABBAABBBAABAABBABBBBBBAABAABAABBABB

*MWG036B *MWG036B BBBBBBAABBAAAABABABABABAABAAABAAAABBBBBBBAAAABBBBABBAAAABBBBBBBAABBAAAABABABABABAABAAABAAAABBBBBBBAAAABBBBABBAAAABABBABBABBBBBBBBBABABABAABBABAAABBAABAABAAAABAABAABBBAABBBABBABBABBBBBBBBBABABABAABBABAAABBAABAABAAAABAABAABBBAABBBAABBAABBBAABAABBABABBBBABBAABABBBABBAABBAABBBAABAABBABABBBBABBAABABBBABB

*RISIC10A *RISIC10A ABBBBBAABBAAAABABABABABAABAAABAAAABBBABABAAAABBBBABBAAAABABBBBBAABBAAAABABABABABAABAAABAAAABBBABABAAAABBBBABBAAAABABBABBABBBBBBBBBABABABAABBABAAABBAABAABAAAABAABAABBBAABBBABBABBABBBBBBBBBABABABAABBABAAABBAABAABAAAABAABAABBBAABBBAABBABBBBAABAABBABABBBBABBAABABBBABBAABBABBBBAABAABBABABBBBABBAABABBBABB

*BCD129 *BCD129 ABBBBBAABBAAABABBABABAAAABAAABAAAABBBABABAAAABBBBABBAAAABABBBBBAABBAAABABBABABAAAABAAABAAAABBBABABAAAABBBBABBAAAABABBABBABBBBBBBBBABBABBABBBBBBBBB--BBBABAABBABAAABBAABAABAAABBBABAABBABAAABBAABAABAAA--BAABAABBAAABBBAABBABBBBAABAABBABBAABAABBAAABBBAABBABBBBAABAABBAB--BBBBABBAABABBBABBBBBBABBAABABBBABB

*ABG320 *ABG320 ABBBBBAABBAAAABABABABAAAABAAABAAAABBBABABAAAABBBBABBAAAABABBBBBAABBAAAABABABABAAAABAAABAAAABBBABABAAAABBBBABBAAAABABBABBABBBBBBBBBABBABBAABBABAAABBAABBABAAAABAABAABBAAABBBABBABBABBBBBBBBBABBABBAABBABAAABBAABBABAAAABAABAABBAAABBBAABBABBBBAABAABBABABBBBBBBAABABBBABBAABBABBBBAABAABBABABBBBBBBAABABBBABB

... ... (Add a few hundred more markers)

A trait datasetA trait dataset^yld1 4.610 5.159 4.701 5.297 4.883 4.681 4.930 4.691 5.411 5.04^yld1 4.610 5.159 4.701 5.297 4.883 4.681 4.930 4.691 5.411 5.041 5.108 5.643 4.924 5.146 4.453 5.031 3.279 5.583 5.488 4.520 4.1 5.108 5.643 4.924 5.146 4.453 5.031 3.279 5.583 5.488 4.520 4.436 5.001 6.275 4.073 436 5.001 6.275 4.073

3.114 4.859 5.401 6.117 4.846 3.860 4.143 5.176 4.735 4.678 4.343.114 4.859 5.401 6.117 4.846 3.860 4.143 5.176 4.735 4.678 4.341 4.651 4.311 5.825 5.929 5.192 4.308 4.567 5.296 5.943 5.747 6.1 4.651 4.311 5.825 5.929 5.192 4.308 4.567 5.296 5.943 5.747 6.010 5.683 5.034 010 5.683 5.034 3.797 5.684 5.872 5.744 4.160 4.489 6.057 5.616 5.273 5.573 4.073.797 5.684 5.872 5.744 4.160 4.489 6.057 5.616 5.273 5.573 4.079 5.192 5.125 3.814 4.833 6.077 5.438 4.073 4.701 5.179 4.702 5.9 5.192 5.125 3.814 4.833 6.077 5.438 4.073 4.701 5.179 4.702 5.731 4.415 5.277 731 4.415 5.277 5.290 4.960 4.136 2.441 5.088 4.967 3.719 4.671 4.970 5.626 5.155.290 4.960 4.136 2.441 5.088 4.967 3.719 4.671 4.970 5.626 5.159 4.459 5.680 4.510 4.039 5.048 4.587 5.381 4.775 5.149 5.798 4.9 4.459 5.680 4.510 4.039 5.048 4.587 5.381 4.775 5.149 5.798 4.281 5.845 4.136 281 5.845 4.136 4.469 5.710 5.626 5.260 4.856 4.382 5.240 4.563 4.362 5.109 4.894.469 5.710 5.626 5.260 4.856 4.382 5.240 4.563 4.362 5.109 4.896 4.099 5.166 5.287 5.199 5.337 3.598 4.641 4.029 5.394 5.804 5.6 4.099 5.166 5.287 5.199 5.337 3.598 4.641 4.029 5.394 5.804 5.862 5.085 4.140 862 5.085 4.140 5.549 4.789 5.509 4.870 3.531 5.213 5.263 4.130 3.195 4.264 3.715.549 4.789 5.509 4.870 3.531 5.213 5.263 4.130 3.195 4.264 3.719 5.421 6.009 5.290 4.907 4.738 5.122 3.931 2.596 5.687 5.219 5.9 5.421 6.009 5.290 4.907 4.738 5.122 3.931 2.596 5.687 5.219 5.031 6.010 4.611 031 6.010 4.611 5.616 3.790 5.371 3.390 5.293 5.522 5.616 3.790 5.371 3.390 5.293 5.522

^yld2 6.829 5.785 6.102 6.022 6.716 5.692 5.544 5.266 7.657 5.71^yld2 6.829 5.785 6.102 6.022 6.716 5.692 5.544 5.266 7.657 5.710 6.516 5.213 7.316 5.687 7.207 4.250 5.447 5.264 6.268 5.044 5.0 6.516 5.213 7.316 5.687 7.207 4.250 5.447 5.264 6.268 5.044 5.512 5.524 6.250 4.886 512 5.524 6.250 4.886 5.128 5.356 4.272 5.945 5.086 5.089 4.863 6.568 5.306 5.602 5.335.128 5.356 4.272 5.945 5.086 5.089 4.863 6.568 5.306 5.602 5.337 5.769 4.375 6.350 6.217 5.342 4.610 4.430 5.449 6.157 4.297 5.7 5.769 4.375 6.350 6.217 5.342 4.610 4.430 5.449 6.157 4.297 5.735 5.431 4.960 735 5.431 4.960 5.567 6.112 6.533 5.266 4.457 6.453 6.570 5.664 5.725 5.336 5.735.567 6.112 6.533 5.266 4.457 6.453 6.570 5.664 5.725 5.336 5.737 5.759 5.461 5.372 5.341 5.091 5.920 6.183 6.277 6.363 5.812 5.7 5.759 5.461 5.372 5.341 5.091 5.920 6.183 6.277 6.363 5.812 5.685 5.544 5.053 685 5.544 5.053 5.501 6.393 4.612 4.360 5.899 5.329 6.015 6.395 5.283 6.636 6.705.501 6.393 4.612 4.360 5.899 5.329 6.015 6.395 5.283 6.636 6.708 4.706 6.425 6.873 5.998 6.123 5.594 5.810 5.376 6.177 5.656 6.8 4.706 6.425 6.873 5.998 6.123 5.594 5.810 5.376 6.177 5.656 6.356 6.050 5.675 356 6.050 5.675 5.286 6.929 6.813 6.851 4.428 4.487 5.967 6.103 4.763 5.777 6.335.286 6.929 6.813 6.851 4.428 4.487 5.967 6.103 4.763 5.777 6.338 6.595 6.561 6.854 5.497 6.640 5.802 6.360 6.358 6.363 6.466 6.8 6.595 6.561 6.854 5.497 6.640 5.802 6.360 6.358 6.363 6.466 6.531 5.193 5.549 531 5.193 5.549 5.381 5.059 7.199 7.755 5.163 4.940 5.196 5.396 5.376 5.932 4.555.381 5.059 7.199 7.755 5.163 4.940 5.196 5.396 5.376 5.932 4.552 7.081 4.986 6.033 5.757 4.806 4.931 5.591 5.909 5.562 6.555 5.2 7.081 4.986 6.033 5.757 4.806 4.931 5.591 5.909 5.562 6.555 5.181 6.323 5.735 181 6.323 5.735 4.417 5.755 5.769 6.406 6.043 5.494 4.417 5.755 5.769 6.406 6.043 5.494

^yld3 2.56 2.07 3.18 2.93 3.07 2.65 3.50 3.03 3.56 4.23 4.27 4.4^yld3 2.56 2.07 3.18 2.93 3.07 2.65 3.50 3.03 3.56 4.23 4.27 4.47 3.65 3.13 2.18 2.63 2.09 3.59 3.53 2.38 2.84 3.43 2.94 2.59 2.7 3.65 3.13 2.18 2.63 2.09 3.59 3.53 2.38 2.84 3.43 2.94 2.59 2.31 3.69 3.51 3.39 3.32 3.90 31 3.69 3.51 3.39 3.32 3.90 2.01 3.55 2.93 3.44 3.51 2.74 3.90 3.53 2.73 3.70 2.91 4.00 3.302.01 3.55 2.93 3.44 3.51 2.74 3.90 3.53 2.73 3.70 2.91 4.00 3.30 3.61 2.75 3.22 3.04 3.57 2.85 3.47 3.67 2.53 3.69 3.98 3.38 2.23.61 2.75 3.22 3.04 3.57 2.85 3.47 3.67 2.53 3.69 3.98 3.38 2.22 3.85 3.26 3.08 3.61 2 3.85 3.26 3.08 3.61 3.00 2.54 4.69 3.39 3.26 2.99 2.65 3.43 3.50 3.30 3.09 3.23 3.713.00 2.54 4.69 3.39 3.26 2.99 2.65 3.43 3.50 3.30 3.09 3.23 3.71 3.57 4.08 1.89 2.86 3.23 2.40 3.47 4.17 4.13 3.66 2.82 3.79 3.73.57 4.08 1.89 2.86 3.23 2.40 3.47 4.17 4.13 3.66 2.82 3.79 3.77 3.71 3.17 3.66 3.59 7 3.71 3.17 3.66 3.59 3.54 2.91 2.45 2.83 4.13 2.53 4.09 4.22 4.42 4.04 4.31 2.85 3.113.54 2.91 2.45 2.83 4.13 2.53 4.09 4.22 4.42 4.04 4.31 2.85 3.11 2.81 3.40 4.52 4.13 3.19 3.02 3.79 3.64 2.79 3.03 3.68 3.52 2.82.81 3.40 4.52 4.13 3.19 3.02 3.79 3.64 2.79 3.03 3.68 3.52 2.86 2.62 2.81 3.62 2.76 6 2.62 2.81 3.62 2.76 2.90 2.43 2.59 3.10 2.26 2.60 2.68 3.62 3.43 3.64 3.65 3.17 3.812.90 2.43 2.59 3.10 2.26 2.60 2.68 3.62 3.43 3.64 3.65 3.17 3.81 2.62 4.33 3.14 3.83 1.60 3.65 3.57 3.85 2.37 4.35 3.62 3.36 3.22.62 4.33 3.14 3.83 1.60 3.65 3.57 3.85 2.37 4.35 3.62 3.36 3.21 3.75 2.46 4.36 3.83 1 3.75 2.46 4.36 3.83

^yld4 3.878 4.554 3.016 2.419 2.952 2.968 3.640 3.495 4.605 3.00^yld4 3.878 4.554 3.016 2.419 2.952 2.968 3.640 3.495 4.605 3.008 3.619 3.239 2.829 3.830 4.594 3.315 3.715 3.085 2.974 2.878 3.8 3.619 3.239 2.829 3.830 4.594 3.315 3.715 3.085 2.974 2.878 3.240 3.749 4.149 3.054 240 3.749 4.149 3.054 3.735 3.186 3.052 3.567 3.790 2.679 3.606 3.945 4.063 3.847 1.743.735 3.186 3.052 3.567 3.790 2.679 3.606 3.945 4.063 3.847 1.748 3.359 2.504 3.430 3.499 3.523 3.358 3.040 3.778 3.183 3.311 3.8 3.359 2.504 3.430 3.499 3.523 3.358 3.040 3.778 3.183 3.311 3.215 3.473 2.071 215 3.473 2.071 3.678 3.604 4.436 3.973 3.356 3.000 3.419 4.161 3.494 4.363 3.723.678 3.604 4.436 3.973 3.356 3.000 3.419 4.161 3.494 4.363 3.729 3.022 3.930 4.054 3.055 4.926 3.636 4.310 3.477 3.725 3.337 4.9 3.022 3.930 4.054 3.055 4.926 3.636 4.310 3.477 3.725 3.337 4.073 3.246 3.878 073 3.246 3.878 3.768 4.295 2.872 3.807 4.141 3.619 3.481 3.007 3.426 3.673 3.603.768 4.295 2.872 3.807 4.141 3.619 3.481 3.007 3.426 3.673 3.604 4.283 3.267 3.438 3.851 4.372 3.980 3.705 3.993 3.993 2.653 3.4 4.283 3.267 3.438 3.851 4.372 3.980 3.705 3.993 3.993 2.653 3.745 3.472 4.416 745 3.472 4.416 3.567 3.226 3.500 4.111 2.936 3.850 4.070 3.509 3.515 2.092 4.123.567 3.226 3.500 4.111 2.936 3.850 4.070 3.509 3.515 2.092 4.120 4.227 4.058 4.221 3.832 3.243 3.332 3.167 3.797 3.188 4.516 4.0 4.227 4.058 4.221 3.832 3.243 3.332 3.167 3.797 3.188 4.516 4.025 3.123 2.987 025 3.123 2.987 3.586 4.318 4.204 3.931 1.510 3.664 3.022 3.544 4.029 3.982 2.643.586 4.318 4.204 3.931 1.510 3.664 3.022 3.544 4.029 3.982 2.645 4.079 3.668 4.079 3.484 3.398 2.713 3.740 2.059 3.603 3.996 3.5 4.079 3.668 4.079 3.484 3.398 2.713 3.740 2.059 3.603 3.996 3.733 4.072 3.538 733 4.072 3.538 3.061 3.495 3.763 4.057 3.886 3.722 3.061 3.495 3.763 4.057 3.886 3.722

^yld5 7.232 6.326 6.533 7.469 5.997 6.525 6.564 7.030 8.432 7.28^yld5 7.232 6.326 6.533 7.469 5.997 6.525 6.564 7.030 8.432 7.287 7.701 6.842 5.735 8.491 6.474 7.289 5.984 7.111 8.048 5.956 6.7 7.701 6.842 5.735 8.491 6.474 7.289 5.984 7.111 8.048 5.956 6.046 6.084 8.435 6.413 046 6.084 8.435 6.413 6.089 7.898 6.781 6.612 6.391 5.821 6.283 7.987 7.473 6.728 5.676.089 7.898 6.781 6.612 6.391 5.821 6.283 7.987 7.473 6.728 5.676 5.590 6.200 5.870 8.052 6.686 7.477 6.348 6.770 8.002 6.983 7.6 5.590 6.200 5.870 8.052 6.686 7.477 6.348 6.770 8.002 6.983 7.381 6.943 5.671 381 6.943 5.671 7.103 7.225 7.141 6.715 6.819 5.612 7.087 6.924 8.096 7.841 7.257.103 7.225 7.141 6.715 6.819 5.612 7.087 6.924 8.096 7.841 7.257 6.882 7.180 6.493 6.672 7.920 7.820 6.441 6.760 6.245 7.185 8.7 6.882 7.180 6.493 6.672 7.920 7.820 6.441 6.760 6.245 7.185 8.163 6.908 6.594 163 6.908 6.594 6.659 7.022 5.711 4.902 7.473 7.381 6.933 6.369 5.931 7.936 7.726.659 7.022 5.711 4.902 7.473 7.381 6.933 6.369 5.931 7.936 7.727 6.709 7.374 6.943 7.510 7.452 6.536 7.513 7.323 6.905 7.306 6.7 6.709 7.374 6.943 7.510 7.452 6.536 7.513 7.323 6.905 7.306 6.454 7.302 6.207 454 7.302 6.207 5.959 7.235 7.088 8.301 6.146 7.810 7.390 7.169 7.348 6.281 7.085.959 7.235 7.088 8.301 6.146 7.810 7.390 7.169 7.348 6.281 7.085 7.390 6.231 6.968 7.471 5.588 7.380 7.287 6.731 7.003 7.001 7.5 7.390 6.231 6.968 7.471 5.588 7.380 7.287 6.731 7.003 7.001 7.539 6.675 6.846 539 6.675 6.846 7.411 6.751 7.117 6.460 7.124 6.788 6.994 7.530 7.329 6.798 5.827.411 6.751 7.117 6.460 7.124 6.788 6.994 7.530 7.329 6.798 5.820 7.798 7.554 6.723 6.806 7.363 5.477 6.391 5.833 7.868 7.189 6.0 7.798 7.554 6.723 6.806 7.363 5.477 6.391 5.833 7.868 7.189 6.017 8.020 7.530 017 8.020 7.530 6.923 6.788 6.592 6.895 6.408 7.005 6.923 6.788 6.592 6.895 6.408 7.005

^yld6 9.2048 7.3174 7.4983 6.8908 6.6689 8.6451 5.6928 5.8669 6.^yld6 9.2048 7.3174 7.4983 6.8908 6.6689 8.6451 5.6928 5.8669 6.7372 8.4881 8.5939 6.7747 6.9113 8.5461 5.4096 8.9283 6.8191 6.67372 8.4881 8.5939 6.7747 6.9113 8.5461 5.4096 8.9283 6.8191 6.6314 7.9488 7.4369 314 7.9488 7.4369 7.8805 7.3106 10.3311 6.7133 7.7065 8.0819 7.1809 9.0307 6.9522 7.8805 7.3106 10.3311 6.7133 7.7065 8.0819 7.1809 9.0307 6.9522 6.8567 7.1638 10.5427 7.3174 4.5256 6.7850 8.3823 6.1195 9.4539 6.8567 7.1638 10.5427 7.3174 4.5256 6.7850 8.3823 6.1195 9.4539 7.6519 7.6519 8.4027 8.7782 9.0614 8.0546 9.3686 5.7474 5.5836 6.4983 7.9522 88.4027 8.7782 9.0614 8.0546 9.3686 5.7474 5.5836 6.4983 7.9522 8.2935 7.4403 6.7577 6.8976 7.8498 8.1945 6.1024 8.2491 8.9556 6..2935 7.4403 6.7577 6.8976 7.8498 8.1945 6.1024 8.2491 8.9556 6.9079 8.4881 9079 8.4881 8.3549 7.1126 6.6655 6.3959 9.5700 7.6143 5.9420 6.7543 7.6416 78.3549 7.1126 6.6655 6.3959 9.5700 7.6143 5.9420 6.7543 7.6416 7.1502 9.7440 6.6212 6.5324 7.2253 8.3993 9.8532 7.1433 6.9932 7..1502 9.7440 6.6212 6.5324 7.2253 8.3993 9.8532 7.1433 6.9932 7.3242 8.0785 3242 8.0785 7.6451 7.5222 7.3584 7.6212 7.1433 7.1604 6.1741 7.0239 5.2048 87.6451 7.5222 7.3584 7.6212 7.1433 7.1604 6.1741 7.0239 5.2048 8.0410 10.3038 7.5973 6.6655 8.3857 6.9761 11.5256 7.1775 6.3003 .0410 10.3038 7.5973 6.6655 8.3857 6.9761 11.5256 7.1775 6.3003 8.4471 8.4471 8.0137 7.5563 7.0683 7.2253 8.5051 7.8328 6.0512 6.9079 8.1331 78.0137 7.5563 7.0683 7.2253 8.5051 7.8328 6.0512 6.9079 8.1331 7.5154 8.0819 7.3481 6.6655 7.7167 7.0205 7.9898 7.4778 9.2594 7..5154 8.0819 7.3481 6.6655 7.7167 7.0205 7.9898 7.4778 9.2594 7.0819 8.0512 0819 8.0512 7.1809 7.5119 8.0000 6.5904 7.2184 7.7201 6.4881 6.7372 6.7645 77.1809 7.5119 8.0000 6.5904 7.2184 7.7201 6.4881 6.7372 6.7645 7.9113 4.5256 8.5495 6.4881 7.7679 9.6280 6.9215 5.1980 7.0171 6..9113 4.5256 8.5495 6.4881 7.7679 9.6280 6.9215 5.1980 7.0171 6.8874 5.3106 8874 5.3106 7.2423 6.9215 6.9590 6.1570 9.5427 7.4915 7.2253 7.2833 6.2218 67.2423 6.9215 6.9590 6.1570 9.5427 7.4915 7.2253 7.2833 6.2218 6.0273 8.0137 7.4642 .0273 8.0137 7.4642

...(...(Add as many more traits as you've got)Add as many more traits as you've got)

7

Marker loci may be mapped by Marker loci may be mapped by multipoint linkage analysismultipoint linkage analysis

Microsatellite map of mouse(http://etd.utmem.edu/WORLD_ACCESS/strom/ABXD5F2markermap.GIF)http://etd.utmem.edu/WORLD_ACCESS/strom/ABXD5F2markermap.GIF)

Relationship between map and Relationship between map and individual genotypeindividual genotype

Probability maps of allele dosage by map position in a single Probability maps of allele dosage by map position in a single individual (2003 thesis of B. Servin; software Grafgene)individual (2003 thesis of B. Servin; software Grafgene)

http://moulon.inra.fr/~fred/servin/thesis/part2http://moulon.inra.fr/~fred/servin/thesis/part2--servin_hospital.pdfservin_hospital.pdf

8

Developing immortal recombinants Developing immortal recombinants for mappingfor mapping

In In inbreedinginbreeding species we can species we can develop develop ““immortalimmortal”” mapping mapping progenies progenies

(RI or DH lines)(RI or DH lines)

We genotype these lines with We genotype these lines with DNA markers for linkage and DNA markers for linkage and QTLQTL mapping.mapping.

Examples: barley, rice, wheat, Examples: barley, rice, wheat, soybean, maize, soybean, maize, ArabidopsisArabidopsis, ..., , ..., mousemouse

Inbredparents

F1

RILsm

arke

r lco

i

self or interbreed

QTL mapping experimentQTL mapping experimentAssay all progeny for genotype and trait phenotypeAssay all progeny for genotype and trait phenotypeSimplest test for a QTL near locus Simplest test for a QTL near locus AA

group progeny by their genotype at locus group progeny by their genotype at locus AAtest for difference in mean of phenotype, using test for difference in mean of phenotype, using tt test or ANOVAtest or ANOVAThe contrast for locus A in the figure would be The contrast for locus A in the figure would be (Y1 + Y4) - (Y2 + Y3)

More elaborate tests are used in practice ("interval mapping")Figure assumes a QTL is near marker A...

what would the plot look like if this were not true?

ABCD

EFG

Y1 Y2 Y3 Y4

mar

ker l

coi

Individual phenotype scores

Y (p

heno

type

val

ue)

Genotype at marker A

(for all RI lines in popn)

aa AA

9

Simplest test for QTL effect in RILsSimplest test for QTL effect in RILs

XP1P2

RI linescolored by (known)genotype at marker A

The estimable QTL effect is the contrast between meanphenotypes of the AA and aa genotype classes

-- ½½½½aaaaAAAA estimates additive effect of substituting A for a

Result: a test statistic such as t, F, LOD, etc.(can obtain by ANOVA or regression)

An FAn F22 design allows testing for design allows testing for dominance too...dominance too...

XP1P2

colored by (known)genotype at marker L

--½½11--½½--110011aaaaAaAaAAAA Write down “orthogonal” contrasts

additive effect (of A allele)

dominance effect

10

Result: a QTL mapResult: a QTL map

Plot test statistic along a genetic mapPlot test statistic along a genetic mapCan use to identify Can use to identify candidate genescandidate genes

having known genetic map or (in sequenced having known genetic map or (in sequenced genome) physical positionsgenome) physical positionsand with annotations suggesting association with and with annotations suggesting association with the traitthe traitSome QTLs have matched known "classical" Some QTLs have matched known "classical" genesgenes

Grupe et al. 2001http://www.sciencemag.org/cgi/content/full/292/5523/1915

Why singleWhy single--marker tests are inferiormarker tests are inferior

We don't want the marker allele effect We don't want the marker allele effect ––we want the QTL allele effectwe want the QTL allele effectMarker effect is less by factor 1 Marker effect is less by factor 1 –– 22rrMQ

farther the marker is from the QTL, the more farther the marker is from the QTL, the more we underestimate the effectwe underestimate the effectEffect size and position are Effect size and position are confoundedconfoundedAlso each mean is of a mixture of Also each mean is of a mixture of QQQQ and and qqqqtrait distributionstrait distributions

calling for a mixture modelcalling for a mixture model

11

Graphically: why oneGraphically: why one--marker test is marker test is misleadingmisleading

Consider QTL Consider QTL Q Q and marker and marker MM, linked as shown in a backcross to , linked as shown in a backcross to genotype genotype mq/mqmq/mq

____M____Q________M____Q____ ____m____q________m____q____ nonrecombinant classesnonrecombinant classes____m____q________m____q____ ____m____q________m____q____ (probability (probability 1 1 -- rrMQMQ))

____M____q________M____q____ ____m____Q________m____Q____ recombinant classesrecombinant classes____m____q________m____q____ ____m____q________m____q____ (probability(probability rrMQMQ))

The effect The effect µµMmMm -- µµmmmm that we can estimate is less than the effect that we can estimate is less than the effect µµQqQq --µµqqqq that we want to estimate, because of that we want to estimate, because of rrMQMQ, the recombination , the recombination probability between marker and QTL.probability between marker and QTL.We can show thatWe can show that

µµMmMm -- µµmmmm = (1 = (1 -- 22rr)()(µµ QqQq -- µµqqqq))so that we always underestimate the contrast by 2so that we always underestimate the contrast by 2rr..

But if the map is dense (a marker every 1 to 5 cM) r is always small and the error is negligible.

QTL "triangulation" QTL "triangulation" –– simple simple interval mapping (SIM)interval mapping (SIM)

Idea: DonIdea: Don’’t test markers; test chromosomal t test markers; test chromosomal positionspositions ("virtual markers")("virtual markers")Regress on Regress on expectation expectation of QTL genotype in an individual, conditioned on of QTL genotype in an individual, conditioned on

genotypes and distances of the two flanking markers.genotypes and distances of the two flanking markers.Consider testing a point Q between markers M1 and M2:Consider testing a point Q between markers M1 and M2:____M1________M1____QQ_________M2_____________M2____

rr11 rr22

Consider the BCConsider the BC11 individual with genotypeindividual with genotype____M1________M1____??_________m2_____________m2________m1________m1____??_________m2_____________m2____

We can write the conditional expectation of the We can write the conditional expectation of the phenotypephenotype asas= = aa * * ff((r1, r2 ) where ) where ff is a nonlinear function, different for each mating is a nonlinear function, different for each mating

design, giving E(design, giving E(ggQQ), the expected QTL genotype of this individual), the expected QTL genotype of this individual

Then we can regress the phenotypes Then we can regress the phenotypes YY on the E(on the E(ggQQ) values. The ) values. The regression coefficient will estimate regression coefficient will estimate aa, the additive effect of a QTL at , the additive effect of a QTL at the tested position.the tested position.

12

Simple interval Simple interval mapping contourmapping contour

7B

6.11

LOD3.0

Xgwm68-7BXksuE18HcXcdo414Xrz508DR_CatDR_Grp94Xfba21Xfbb189aDR_Cht1b-7BDR_Tha1-7BDR_Tha2-7BXksuD2aXfbb175Xfba305Xwg686Xcdo686Xwg514Xrz476Xbcd178Xabc455Xbcd349Xbcd98XbXcdo551Xbcd98XaXbcd1338Xbcd385Xwg180Xbcd310Xfbb150Xgwm573-7BDR_Cbp2-7BDR_Pr1b-7BXfba42b

0.0

Interval analysis for traitsLR93.1_Ithaca LR93.2_IthacaLr96_Ithaca

Plot regression Plot regression FFtest, LOD statistic, test, LOD statistic, etc.etc.Leaf rust severity in Leaf rust severity in wheat, 3 dates in 2 wheat, 3 dates in 2 yearsyears...note candidate ...note candidate DR genes!DR genes!

Why SIM is wrong tooWhy SIM is wrong tooThe biggest flaw of SIM:The biggest flaw of SIM:

The hypothesis test is ofThe hypothesis test is of

HH11: a single QTL at this position, : a single QTL at this position, vs.vs.HH00: no QTL anywhere.: no QTL anywhere.

But that assumes there's only ONE QTL for the trait!But that assumes there's only ONE QTL for the trait!...which is silly, if we know up front that it's probably a poly...which is silly, if we know up front that it's probably a polygenically genically

controlled trait.controlled trait.

We really need to test a multipleWe really need to test a multiple--QTL model, QTL model, HH11: including...: including... (the full model) vs.vs.HH00: excluding...: excluding... (the reduced model)

a QTL at the tested position.a QTL at the tested position.

13

What we might miss with SIMWhat we might miss with SIM

What are potential consequences of using What are potential consequences of using an inappropriate test?an inappropriate test?

Ghost QTLsGhost QTLs(see single QTL peak where there should be two (see single QTL peak where there should be two with effects from same parent)with effects from same parent)

Invisible QTLsInvisible QTLsfail to see QTLs where there are two linked ones fail to see QTLs where there are two linked ones with opposite effectswith opposite effects

Do these errors happen all the time?Do these errors happen all the time?No, probably rarely.No, probably rarely.

Composite interval analysis (CIM)Composite interval analysis (CIM)

A currently popular QTL mapping method, A currently popular QTL mapping method, CComposite omposite IInterval nterval AAnalysis or nalysis or CIMCIM, , employs a multipleemploys a multiple--regression model or regression model or the maximumthe maximum--likelihood equivalent.likelihood equivalent.The statistical model:The statistical model:

where the where the bb are effects of selected are effects of selected "background" markers selected to absorb "background" markers selected to absorb effects of unlinked QTLseffects of unlinked QTLs

Yi = µ + aE(G*) + ∑=

B

kikk Gb

1

. + ei

14

CIM vs. SIM outputCIM vs. SIM output

Credit: whoever I stole this slide from in 2001

Many more methods & variantsMany more methods & variantsMIM MIM (Zeng & colleagues, NCSU)(Zeng & colleagues, NCSU)

Multiple interval mappingMultiple interval mappingreplaces all effects with QTLsreplaces all effects with QTLs

Multiple trait analysis Multiple trait analysis (Jiang et al., Korol et al., since 1995)(Jiang et al., Korol et al., since 1995)uses correlated traits to increase power of QTL testsuses correlated traits to increase power of QTL tests

Bayesian and pseudoBayesian and pseudo--Bayesian analysesBayesian analysesMany authorsMany authors

Combined analysesCombined analysesover multiple environmentsover multiple environmentsestimating main, environmental, QTL x E, QTL x QTL, etc.estimating main, environmental, QTL x E, QTL x QTL, etc.

LinkageLinkage--disequilibrium mappingdisequilibrium mappingin experiments that don't employ planned crossesin experiments that don't employ planned crosses

15

Software to useSoftware to use

Of noncommercial software, probably QTL Of noncommercial software, probably QTL Cartographer (Wang, Basten, Zeng et al.) most Cartographer (Wang, Basten, Zeng et al.) most commonly reported and most complete packagecommonly reported and most complete packageMany other free applicationsMany other free applications

in general a bit tedious to usein general a bit tedious to useand lack many exploratory and/or visual featuresand lack many exploratory and/or visual features

Best commercial: MapQTL and MultiQTLBest commercial: MapQTL and MultiQTL... don't ask... don't ask

But how do you know it's a QTL? But how do you know it's a QTL? Data exploration & horror storiesData exploration & horror stories

GIGO is the ruleGIGO is the ruleKnow your data quality and behaviorKnow your data quality and behaviorDon't trust anybodyDon't trust anybody

well, not blindly anywaywell, not blindly anyway

Case studiesCase studies1.1. Synthetic Synthetic xx cultivated wheat BCcultivated wheat BC22FF44 popnpopn2.2. Wheat RI popn: two different studiesWheat RI popn: two different studies

16

Case study ICase study ISynthetic Synthetic xx cultivated wheat BCcultivated wheat BC22FF44 popnpopn

Purpose of "advanced backcross" study was to find Purpose of "advanced backcross" study was to find QTLs from wild wheat (QTLs from wild wheat (Aegilops tauschii).Aegilops tauschii).189 lines were developed, genotyped with 153 SSR 189 lines were developed, genotyped with 153 SSR markersmarkersYieldYield--related traits scored in two environmentsrelated traits scored in two environmentsLinkage map couldn't be built with conventional Linkage map couldn't be built with conventional softwaresoftware

(mating design is nonstandard)(mating design is nonstandard)so wellso well--known reference map was usedknown reference map was used

Analysis was by singleAnalysis was by single--marker regressionmarker regressionwith attempt to use simple interval analysis toowith attempt to use simple interval analysis too

Case study I: The horror beginsCase study I: The horror beginsStudent reported about 50 QTLsStudent reported about 50 QTLs

based on a nicebased on a nice--sounding sounding PP value threshold of 0.001value threshold of 0.001sample QTL scan looks like this:sample QTL scan looks like this:

1A1B 1D 2A 2B 2D 3A 3B 3D 4A 4B 4D 5A 5B 5D 6A6B6D 7A7B 7D

Regression for trait HD_Manh, dubious data

Karl

SynF 0.0

12.1

12.1

Marker loci on map

CN's rule of thumb for bump hunters:CN's rule of thumb for bump hunters:If you see a lot of "QTLs" with effects barely above background,If you see a lot of "QTLs" with effects barely above background,they're probably bogus.they're probably bogus.

Don't put all your money on Don't put all your money on PP values or other stats, but...values or other stats, but...Get more dataGet more dataGet better dataGet better dataCheckCheck your data, experimental design, mapyour data, experimental design, map

17

Case study I: When to accept a QTL?Case study I: When to accept a QTL?

Here, few QTL test statistics barely exceed 95% of the Here, few QTL test statistics barely exceed 95% of the values reached with randomized datavalues reached with randomized data

CN does permutation tests for individual CN does permutation tests for individual trait/chromosome combinationstrait/chromosome combinations

scramble trait data, rerun test; do this 1000 timesscramble trait data, rerun test; do this 1000 timeshow does your original stat look against this distribution of sthow does your original stat look against this distribution of stats?ats?

1A1B 1D 2A 2B 2D 3A 3B 3D 4A4B4D5A5B 5D 6A6B6D 7A7B7D

Permutation analysis for HD_Manh, chromosome 3B Shuffles 1000

Marker loci on map

5.0

10.3

LOD

95%ile99%ileMax

Case study I: Permutation tests, IICase study I: Permutation tests, II

For a trait controlled by one major geneFor a trait controlled by one major gene

1A1B 1D 2A 2B 2D 3A 3B 3D 4A4B4D5A5B 5D 6A6B6D 7A7B7D

Permutation analysis for HARD_Manh, Shuffles 1000

Marker loci on map

14.5

9.2

LOD

95%ile99%ileMax

One QTL test statistic far exceeds all of the values One QTL test statistic far exceeds all of the values reached with randomized data; the other, maybe 95%reached with randomized data; the other, maybe 95%

18

Case study I: Knocked 50 QTLs Case study I: Knocked 50 QTLs down to a dozendown to a dozen

Take a look at the trait distributionsTake a look at the trait distributionsnotice one wheat line contains 44.8% proteinnotice one wheat line contains 44.8% proteinNot likely. There goes a protein "QTL"Not likely. There goes a protein "QTL"

17.2 24.1 31.0 37.9 44.8

50

100

PROT_Hutch

Case study I: how was map made?Case study I: how was map made?

Data and map both contained markers with Data and map both contained markers with multiple fragments multiple fragments

123456

789

1011

121314

151617

29391

35

52

121

14

62

20

42

Chr2 QYld.ksu-2D

QTw.ksu-2D

MarkerMarker 6 has a locus on chrom 2 of has a locus on chrom 2 of reference map and another locus reference map and another locus on a different chromosome.on a different chromosome.Student, seeing two fragments, Student, seeing two fragments, scored them as different loci.scored them as different loci.

But they're different alleles of the But they're different alleles of the same locus!same locus!

Student randomly assigned one Student randomly assigned one "locus" to chrom 2, the other to the "locus" to chrom 2, the other to the other chromother chromBecause the marker's associated Because the marker's associated with a QTL, student reported two with a QTL, student reported two QTLs, one on each chrom.QTLs, one on each chrom.Dump another QTL

19

Case study I: Map was a bit sparse..Case study I: Map was a bit sparse..

Some chromosomes had many markers, Some chromosomes had many markers, others only two or three.others only two or three.

Why? How?Why? How?

Case study I: how was popn made?Case study I: how was popn made?

Student made 15 BCStudent made 15 BC11FF11Backcrossed each, developed BCBackcrossed each, developed BC22FF11 families of families of 15 to 20 plants15 to 20 plantsSelected against shattering, tight glumes, etc. in Selected against shattering, tight glumes, etc. in greenhousegreenhouseEventually selected 480 of 1200 BCEventually selected 480 of 1200 BC22FF33 in in greenhousegreenhouseGrew in field, selected 189 BCGrew in field, selected 189 BC22FF44 lines for lines for mapping.mapping.Problems with this?Problems with this?

20

Case study I: selection and popn Case study I: selection and popn developmentdevelopment

Take a look at the breeding notes:Take a look at the breeding notes:

From how many BCFrom how many BC11FF11 parents did the BCparents did the BC22 families actually families actually descend?descend?

8 of the original 15.8 of the original 15.Hmm...Hmm...

From how many BCFrom how many BC11FF11 parents did the final BCparents did the final BC22FF44 lines actually lines actually descend?descend?

3 of the 83 of the 8..Were they uniformly represented?Were they uniformly represented?

No. No. A single BCA single BC11FF11 plant accounted for plant accounted for 164 of the of the 189 lines.lines.Why?Why?

SelectionSelection..Consequences?Consequences?

A lot of genome was already fixed to recurrent parent alleles atA lot of genome was already fixed to recurrent parent alleles at BCBC11That part is invisible to any kind of genetic mapping. That part is invisible to any kind of genetic mapping. It has no allelic It has no allelic variation!variation!

Case study II: Is the map right?Case study II: Is the map right?

Widely used wheat RIL popn, 114 linesWidely used wheat RIL popn, 114 linesUsed to map genes for root nematode Used to map genes for root nematode tolerance, with data collected in two yearstolerance, with data collected in two yearsA student used the published mapA student used the published map

right out of the public online database.right out of the public online database....why not? Can't you trust a published map?...why not? Can't you trust a published map?

21

Case study II: Is the map right?Case study II: Is the map right?CIM scans using original map:CIM scans using original map:

Xgbx3165

Xpsr889 Xbcd1821 Xbcd342 Xabg466

Xbarc183 Xbarc173

Xpsr964 Xfba307

Xgwm469

LRS 19.0

45

0 cM

CIM scans using recomputed map:CIM scans using recomputed map:

Case study III: What kind of trait is Case study III: What kind of trait is this?this?

In same RIL set as case II, a complex biochemical trait In same RIL set as case II, a complex biochemical trait was tested.was tested.Data looked like this:Data looked like this:

00 7575 100100 12.512.5 00 5050 2525 7575 37.537.5 10010087.587.5 5050 12.512.5 87.587.5 12.512.5 62.562.5 00 37.537.500 00 87.587.5 12.512.5 87.587.5 12.512.5 2525 0062.562.5 7575 12.512.5 2525 37.537.5 12.512.5 2525 37.537.512.512.5 37.537.5 87.587.5 7575 62.562.5 00 62.562.5 005050 00 00 37.537.5 00 .. 00 10010037.537.5 00 2525 2525 00 00 87.587.5 37.537.52525 7575 62.562.5 00 2525 12.512.5 00 12.512.500 7575 2525 00 00 62.562.5 100100 252500 2525 00 100100 5050 00 37.537.5 0012.512.5 00 2525 100100 12.512.5 7575 7575 252512.512.5 37.537.5 7575 7575 5050 37.537.5 87.587.5 12.512.52525 5050 12.512.5 12.512.5 5050 62.562.5 12.512.5 ..

.. 5050 62.562.5 00 .. .. 2525 12.512.5

QTL effects looked suggestive for some markers QTL effects looked suggestive for some markers evaluated on 50 linesevaluated on 50 lines

...but evaporated when rest of lines were genotyped....but evaporated when rest of lines were genotyped.Why is this trait suspicious?Why is this trait suspicious?

22

SummarySummary

QTL stats are interesting, deep, and cool, butQTL stats are interesting, deep, and cool, butquite basic ones will take you far.quite basic ones will take you far.However...However...

To do meaningful QTL mapping:To do meaningful QTL mapping:Know the basics of genetics and statisticsKnow the basics of genetics and statisticsKnow your dataKnow your dataKnow the effects of errors on your stats resultsKnow the effects of errors on your stats resultsBe suspicious!Be suspicious!