Empirical Analysis of Selection Screens for Domestication and Improvement Loci in Maize by Extended DNA Sequencing
- Masanori Yamasaki,
- Steven G. Schroeder,
- Hector Sanchez-Villeda,
- Brandon S. Gaut, and
- Michael D. McMullen*
- Masanori Yamasaki, Steven G. Schroeder, Hector Sanchez-Villeda, and Michael D. McMullen, Division of Plant Sciences, Univ. of Missouri, Columbia, MO 65211; Brandon S. Gaut, Dep. of Ecology and Evolutionary Biology, Univ. of California, Irvine, CA 92697; Michael D. McMullen, USDA-ARS, Plant Genetics Research Unit, Columbia, MO 65211. Masanori Yamasaki, present address: Food Resources Education and Research Center, Kobe Univ., Kasai, Hyogo 675-2103, Japan. Steven G. Schroeder, Present address: USDA-ARS, Bovine Functional Genomics Lab., Beltsville, MD 20705.
Abstract
Both domestication and crop improvement in maize (Zea mays ssp. mays) have involved selection of specific alleles at genes controlling key morphologic and agronomic traits, resulting in reduced genetic diversity relative to unselected genes. This difference in genetic diversity has led to the development of genomic screens for artificial selection in maize that have identified ∼50 candidate agronomic genes. One limitation of these initial genome screens is that the short length of the alignment (average length < 300 bp) restricts the statistical power and may lead to false positives. The major objective of this research was to provide an empirical determination of the false positive rate of genomic screens for artificial selection in maize. Therefore, we performed extended sequencing throughout the available gene sequence of 27 previously identified selection candidates using maize inbred, maize landrace (for 12 genes), and teosinte (Zea mays ssp. parviglumis) accessions. The extended sequence alignments (average length > 2000 bp) allowed clear separation of strong candidates for selection from those that cannot be distinguished from the tail of the diversity distribution of all maize genes. The extended alignments also allowed linkage disequilibrium to be considered in evaluating a candidate's selection status. This proved particularly useful in distinguishing selection at domestication versus subsequent crop improvement.
Among the key events in the advancement of human civilization was the domestication of major cereal crop species from wild progenitor populations. The domestication events resulted in the original landrace varieties, which were adapted to a wide range of environmental conditions. Subsequently, the landraces provided the genetic material for modern plant breeders to select improved varieties and inbred lines by enhancing traits controlling agricultural productivity and performance. It is logical to assume that genes that have been targets of artificial selection during domestication and plant breeding are significant loci in determining agronomic traits (Vigouroux et al., 2002; Ross-Ibarra et al., 2007).
Maize (Zea mays ssp. mays) was domesticated from teosinte (Zea mays ssp. parviglumis) through a single domestication event in southern Mexico between 6000 and 9000 years ago (Piperno and Flannery, 2001; Matsuoka et al., 2002). Despite its selection history, the majority of maize genes exhibit high levels of nucleotide diversity and low linkage disequilibrium (LD) (Tenaillon et al., 2001). The difference in genetic diversity in genes that are targets of selection compared to an unselected control set of genes has permitted the development of population genetics approaches to identify selection candidates in maize (Vigouroux et al., 2002; Wright et al., 2005; Yamasaki et al., 2005). These approaches are known as selection screens or selective sweep mapping and represent an unbiased, “bottom-up” approach to identifying genes of agronomic importance (Ross-Ibarra et al., 2007; Yamasaki et al., 2007). Using a selective sweep strategy that contrasted diversity in maize to diversity in teosinte, Wright et al. (2005) concluded that 2 to 4%, or ∼1200 maize genes exhibit evidence of artificial selection. In an alternative approach, Yamasaki et al. (2005) proposed an efficient screen for the identification of potentially selected maize genes by focusing on genes with zero single nucleotide polymorphism (SNP) diversity in maize inbreds. Genes strongly impacted by selection are expected to be enriched in the subset of genes that exhibit very low genetic diversity (Teshima et al., 2006).
To date, maize has been the most successful species for conducting selection screens in crop plants. Attempts to define selected genes in sorghum (Sorghum bicolor L. Moench) (Casa et al., 2006; Hamblin et al., 2006) have not been as successful, possibly due to reduced power from lower overall levels of sequence diversity and frequency distributions indicative of a complex demographic history for cultivated sorghum. Barley (Hordeum vulgare L. Poaceae) (Morrell et al., 2005) and sunflower (Helianthus annuus L.) (Liu and Burke, 2006) have high levels of diversity in progenitors that make them promising candidates for identifying genes of potential agronomic importance via selection screens. Characterization of selection in animal species has been most extensive for Drosophila species and for humans (Harr et al., 2002; Kayser et al., 2003; Jensen et al., 2005). Major differences in demographic histories of these species compared to maize require species-specific models for evaluating effectiveness of selection screens (Teshima et al., 2006).
In both Wright et al. (2005), and Yamasaki et al. (2005), the selective sweep mapping in maize involved sequencing a PCR amplification product from each gene (average length: ∼300 bp), with polymorphism based on a sample size (n) of less than fourteen maize individuals. The statistical power of selective screens may be quite limited with this sampling design, suggesting that many selected genes were not detected—i.e., there was a high rate of false negatives (Type II error). More importantly for the current study, selection screens typically identify loci in the tail of the genomic distribution of polymorphism, yet it is possible that non-selected genes can also be found in the tails of the distribution. In other words, there are undoubtedly false positives, but the level of Type I error is difficult to quantify precisely. It is thus appropriate to attempt to verify inferences based on the genomic selection screen. To this end, Yamasaki et al. (2005) selected eight selection candidates with strong evidence of selection, and performed extended sequencing throughout the available gene sequence. The rationale was that strongly selected genes should have both low diversity in maize and big differences in diversity between maize and teosinte throughout the gene. However, because only the eight most promising candidates were analyzed the general false-positive rate for selective sweep mapping in maize remained unclear.
The major objective of this paper is to estimate the false-positive rate of genomic screens for artificial selection in maize when the initial sequencing has been performed on a single, relatively short sampling of a gene (∼300 bp) as was conducted in our previous studies. In this paper a total of the 27 selection candidates are analyzed by extended sequencing to empirically examine the potential false-positive rate of selective sweep mapping. In addition, genes from Yamasaki et al. (2005) that were demonstrated to be selected between inbreds and teosintes were also sequenced in a panel of landrace accessions to determine whether selection on the candidate genes occurred primarily during domestication or crop improvement.
Materials and Methods
Plant Materials and Extended DNA Sequence Analysis for Selected Candidate Genes
Three diverse sets of maize materials were used for DNA sequence analysis, maize inbred lines, maize landraces, and partially inbred teosinte accessions (Supplementary Table 1, Yamasaki et al., 2005). For maize inbreds, we used either of two sets, the same 14 diverse lines as Wright et al. (2005) and Yamasaki et al. (2005) or an expanded set of 28 diverse lines. We conducted extended DNA sequencing throughout the available gene sequence of a total of 27 selected candidate genes identified by either of two independent tests, coalescent simulation (CS) or Hudson-Kreitman-Aguadé (HKA) (Yamasaki et al., 2005; Wright et al., 2005). The sets of PCR primers to analyze throughout the selected candidate genes were designed using the Primer3 program (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi; verified 29 April 2008). The genomic PCR was performed with PCR Master Mix (Promega, Madison, WI) or TaKaRa LA Taq with GC buffer (TAKARA BIO INC., Otsu, Japan) using DNA Engine Tetrad thermocycler (MJ Research, Watertown, MA) (one cycle of 1 to 2 min at 95°C; 35 cycles of 1 min at 95°C, 1 min at 50 to 65°C [dependent on primers] and 30 sec to 6 min [dependent on PCR product] at 72°C; final extension of 3 to 10 min at 72°C). Following PCR amplification, unincorporated primers and deoxynucleotide triphosphates were removed by exonuclease I (New England Biolabs, Ipswich, MA) and shrimp alkaline phosphatase (United States Biochemical, Cleveland, OH) and products were ethanol precipitated before sequencing. The PCR products were sequenced with forward, reverse, and internal primers using BigDye terminator version 3.1 (Applied Biosystems, Foster City, CA) and analyzed on ABI 3100 sequencer (Applied Biosystems). Base calling, quality assessment, and trimming of trace files were conducted with PHRED (Ewing and Green, 1998; Ewing et al., 1998) and sequence assembly was performed by PHRAP. The multiple sequences for each gene were aligned with ClustalW (Thompson et al., 1994), and edited with an in-house sequence aligning program, DNAAlignEditor (Sanchez-Villeda et al., 2008) and Sequencher version 4.6 (Gene Codes Corporation, Ann Arbor, MI). Sequences were removed if they had an average PHRAP quality score <30 or were <80% of the average length of the sequences in the alignment. Only SNP variants with a PHRAP quality score >30 were used for analysis. All alignments are available from the Maize Diversity project website, www.panzea.org (verified 28 April 2008). All DNA sequences are deposited in GenBank under the accession numbers BV725508–BV726468.
Population Genetic Analysis and Coalescent Simulation
Population genetics parameters for the selected candidate genes were measured for each population (maize inbreds, maize landraces, and teosintes) in DnaSP version 4.10 (Rozas et al., 2003). The number of polymorphic sites, S, the number of unique sequences (haplotypes), h, the average proportion of pairwise nucleotide differences per nucleotide site, π (Tajima, 1983), and LD parameter r 2 (Hill and Robertson, 1968) was determined for each population for each gene.
We used CS to model the impact of a bottleneck on sequence diversity using protocols detailed previously (Tenaillon et al., 2004; Wright et al., 2005; Yamasaki et al., 2005) (Supplementary Fig. 1). The method first estimates the demographic history of reference genes. Two demographic scenarios were estimated separately—a “domestication” bottleneck based on comparisons between teosinte and landrace data, and an “improvement” bottleneck based on comparisons between teosinte and inbred data. We fixed t, d, N a, N inbred, and N landrace based on previous simulation studies (Eyre-Walker et al., 1998; Tenaillon et al., 2004; Wright et al., 2005). The time of the split between maize and teosinte t and the duration of the maize population bottleneck d were assumed to be constant across loci and set to 7500 and 2800 generations, respectively. The ancestral population (teosinte) size N a and the population size of maize inbreds and maize landraces N inbred and N landrace were assumed to be 940,000 and 1,000,000, respectively. The mutation rate μ (Watterson, 1975) and recombination parameter ρ (Hudson, 2001) were estimated from teosinte data. The maximum likelihood (ML) of the ratio k (= N b/d) of bottleneck duration and strength was previously estimated with data from four neutral genes [adh1, bz2, fus6, and glb1 (Eyre-Walker et al., 1998; Hilton and Gaut, 1998; Tenaillon et al., 2001)]. For the landraces, the estimated neutral multilocus parameter was k = 4.65, and for the inbreds the estimated parameter was k = 1.25 as previously reported by Yamasaki et al. (2005). For all simulations, the goodness-of-fit statistic was the number of segregating sites in maize. The CS tests whether the candidate gene has a history concordant with the neutral loci at P = 0.05 (Innan and Kim, 2004).
Results and Discussion
DNA Sequence Diversity in Selected Candidate Genes
Using maize inbred lines and partially inbred teosinte accessions, we conducted DNA sequencing throughout the available gene sequence of a total of 27 genes (Table 1). These included all 17 candidates identified as significant for selection by Yamasaki et al. (2005) by either HKA (Hudson et al., 1987) or CS of domestication analysis, or both. In addition, 10 of the top 20 candidates from Wright et al. (2005) were analyzed. These candidates were ranked 2, 3, 5, 8 11, 12, 15, 17, 18, and 20 in order of significance by the posterior probability that they should be assigned to the selected class of genes, and their posterior probabilities ranged from 0.65 to 0.43 (Wright et al., 2005). For the apparently selected genes from Yamasaki et al. (2005) (see below), we also sequenced the gene in a diverse panel of maize landraces to investigate whether the loss of diversity occurred primarily during domestication or more recently during crop improvement (Table 1, Fig. 1). The functional classifications of the proteins encoded by these candidate loci were determined by nucleotide and protein homology searches and are given in Wright et al. (2005) and Yamasaki et al. (2005).
Sequence diversity in maize inbreds, landraces, and teosintes for extended sequencing in candidate selected genes.
Sliding-window analysis of the genetic diversity π of candidate genes. For sliding-window analysis, π was calculated for segments of 100 bp at 10 bp intervals. Horizontal axis on the graphs indicate DNA sequence position whereas vertical axes means genetic diversity, π. Red, green and blue lines indicate genetic diversity in the inbreds, landraces and teosintes, respectively. For gene structure under the sliding-window graphs, white bars indicate the predicted exons and black lines indicate introns or genomic regions. Left-arrows and right-arrows indicate the positions for predicted start codons and stop codons, respectively. The lines with two arrows under the gene structure indicate the sequenced regions in our initial screening. A. Genes significant for selection throughout the entire gene sequence. B. Genes that were not significant for selection throughout the entire gene sequence. C. Neutral (control) gene sequences.
Extended DNA sequencing of these 27 candidates yielded approximately seven-fold longer average lengths of alignment (without gaps) than the original selection screens. The average length of the resulting alignments were 2122 nucleotides (nt) in inbreds and 2018 nt in teosintes (Table 1), as compared to 276 nt in inbreds and 298 nt in teosinte for previous studies (Yamasaki et al., 2005; Wright et al., 2005). Over all 27 genes, there was a significant decrease in the nucleotide diversity measure π (Tajima, 1983) and the number of segregating sites (S) between maize inbreds and teosintes (Wilcoxon signed-rank test, both P < 0.001). We also obtained extended nucleotide sequence for 12 candidates using maize landraces, averaging 2211 nt without gaps (Table 1, Fig. 1). The diversity measures π and S in the landrace samples exhibited values intermediate to maize inbreds and teosintes as expected from their historic position being derived from teosintes and precursor to inbreds (Table 1).
Confirmation of Selection Candidates by Extended Sequencing
CS analysis was used to test for selection in the extended sequence of the candidate genes previously reported by Yamasaki et al. (2005) (Table 2). The CS analyses incorporate summary statistics, including information about recombination, to estimate the duration and severity of the bottleneck (Supplementary Fig. 1). The CS tests whether the loss of diversity in inbreds versus teosintes at a candidate locus is too great to be explained by demographic effects alone (Tenaillon et al., 2004; Wright et al., 2005; Yamasaki et al., 2005). Because the CS analysis accounts for the expected general diversity loss from bottleneck effects, the CS analysis is a potentially powerful test for artificial selection. We define a ‘selected gene’ as a gene that deviates significantly from the null demographic model, based on the CS test, in the inbred versus teosinte comparison. Eleven of the 17 candidate loci were significant for the entire gene sequence by this criterion (Table 2). Visual inspection of the distribution of the genetic diversity of inbred and teosinte lines across the genes as illustrated by a sliding window display of π shows two distinct patterns (Fig. 1). Either the reduced diversity in the inbreds is evident throughout the gene, or it is restricted to the region of the initial sequence. The only two exceptions to these patterns are AY106616 and AY109011. AY106616 is not significant across the entire gene, but as previously noted by Yamasaki et al. (2005), a diversity region of 1.5 kb at the 3′ end of the gene suggests that selection affected only a portion of the gene sequence. AY109011 was significant for the entire gene sequence, but examination of the π distribution suggests that the 3′ 2 kb of the gene has diversity patterns consistent with selection, but the 5′ 1 kb has a pattern of genetic diversity similar to an unselected gene. Eleven of 17 genes selected as candidates based on a single, short region were significant for the entire gene sequence, supporting selection for 65% of our previous inferences.
Tests of selection in maize inbreds, maize landraces and teosintes in candidate selected genes from Yamasaki et al. (2005).
Yamasaki et al. (2005) previously confirmed the candidate status of seven of eight of the genes in this dataset. These were the eight candidates identified to have diversity patterns consistent with selection by two different tests, HKA and CS. For the remaining nine candidates, initially detected as significant by only one test or ambiguous with regard to assignment as domestication or improvement, our current data are consistent with selection for five of nine. Although the sample sizes are small it appears that selective screen candidates significant for selection by both HKA and CS are more concordant with inferences based on full gene sequences (less apparent false positives) than inferences from only a single test. The HKA test relies on divergence information relative to an outgroup species (Hudson et al., 1987), whereas CS specifically incorporates the demographic history and primarily tests loss of diversity from progenitor to domesticate. As these two analyses test different expectations of artificial selection it is not surprising that genes in selective screens that pass both tests have the highest support for candidate selection status.
CS was also used to test for selection in the extended sequence of 10 candidate genes from Wright et al. (2005) (Table 3 and Fig. 2). Four of the 10 candidates (AY107228, AY110082, AY107907, and AY107903) deviated significantly from the null model throughout the available gene sequence. Since the posterior probabilities estimated by Wright et al. (2005) averaged 53% for the top 20 genes, our extended sequencing yields an apparent false positive rate that is consistent with their calculations. Extrapolating from their original analyses, Wright et al. (2005) estimated that 2 to 4% of maize genes (or ∼1200 to 2400 genes) contained a signature of selection. If we accept that 50% confirmation rate as suggested by our extended sequencing data combined across the two data sets, this implies that 1 to 2%, or 600 to 1200, genes in maize will have long regions of sequence with reduced diversity in modern maize lines consistent with those genes having been targets of artificial selection.
Tests of selection in maize inbreds, maize landraces and teosintes in candidate selected genes from Wright et al. (2005).
Sliding-window analysis of the genetic diversity π of candidate genes initially identified by Wright et al. (2005). Graphs are identical in form to Fig. 1. A. Genes significant for selection throughout the entire gene sequence. B. Genes not significant for selection throughout the entire gene sequence.
Recently, computer simulations have greatly enhanced our understanding of the statistical properties of genomic screens for selection. For example, Teshima et al. (2006) simulated sequence evolution based on the bottleneck model inferred by Wright et al. (2005). The simulations included unselected genes as well as genes selected during domestication. The goal was to determine how easily selected genes could be identified by empirically ranking summary statistics like π and identifying the extreme genes as candidate selection genes. The results depend on the assumptions of the simulations, such as the strength of selection [which at 5%, may have been conservative (Olsen et al., 2006)], the mode of selection (i.e., whether selection targeted new or old mutations, and the level of dominance), the sequence length (5 kb), and the proportion of selected to nonselected genes. The general result was that the tails of the distributions of summary statistics were highly enriched for selected genes when selection had acted on a new mutation. Rates of false positives were low but false-negative rates were still appreciable (Teshima et al., 2006). Nonetheless, given the high variability in maize and the bottleneck history of maize, the simulations are consistent with the empirical studies in that they strongly suggest that selection screens should yield reasonable candidates.
The pressing question is how one confirms selection in candidates from an initial screen. The approach we have taken is to perform more sequencing and, for 12 genes, sequence a different panel of individuals (see below). We must admit, however, to an intrinsic ascertainment bias with this approach. Briefly, if a genic region has low diversity in an initial screen based on short sequence fragments, that genic region is more likely to reside in an extended region of low diversity. This is true whether the fragment is selected or not (Thornton and Jensen, 2007). Thus, it is possible that some of our candidate genes with low diversity throughout the gene have not been affected by selection. Extended sequencing is still valuable, however, as it continues to refine our list of ‘best candidates,’ along with our best estimates of the proportion of false-positives.
Timing of Selection: Domestication versus Crop Improvement
Yamasaki et al. (2005) used a panel of maize landrace accessions to divide selection in maize into domestication and crop improvement. Hufford et al. (2007) conducted a similar analysis of selection candidates from Wright et al. (2005). In both studies a candidate gene is considered a ‘domestication’ gene if the gene deviated from the null model in the contrast between both landraces versus teosintes and inbreds versus teosintes. A gene is inferred to be an ‘improvement’ gene if it is significant in inbreds versus teosintes, but not landraces versus teosintes. On the basis of the short initial sequences in inbreds, landraces, and teosintes about an equal number of selected genes were categorized as domestication versus improvement for both of the prior selection scans (Yamasaki et al., 2005; Hufford et al., 2007).
In addition to strengthening support for selection we reasoned that extended sequencing can also be useful in clarifying the timing of selection. We therefore performed extended sequencing in landrace samples for the 12 genes from Yamasaki et al. (2005) (Table 2, Fig. 1). This included the 11 genes selected throughout the entire gene sequence and AY106616, which shows selection over 1.5 kb. Coalescent simulation of domestication comparing inbreds versus teosintes and landraces versus teosintes were used to assign the 12 genes as domestication or crop improvement genes. From this analysis nine of the 12 genes were assigned as domestication genes (Table 2). For genes AY107195, AY110109, AY108178, and AY107821 the analysis of the extended alignments changed the classification from improvement (from Yamasaki et al., 2005) to domestication (Table 2). This classification scheme admittedly assumes that the selection process is discrete and bimodal, but for many genes it may have been a continuous process. Moreover, our approach assumes that the landrace and teosinte data are historical representations of genetic diversity, but both samples are contemporaneous. Nonetheless, it appears from our extended sequencing data that the majority of the genes identified as selection candidates were targeted before the formation of the landraces.
In addition to the loss of sequence diversity another expected genetic consequence of artificial selection is that targets of selection will exhibit increased LD relative to unselected genes (Palaisa et al., 2004). In maize, LD generally decays very rapidly in neutral genes in all diverse population sets: teosintes, landraces, and inbreds (Remington et al., 2001; Tenaillon et al., 2001) (Table 2). The length of the extended sequencing alignments permits LD to be considered as a supporting evidence for selection. From the LD plotting functions of DnaSP (Rozas et al., 2003) we estimated the correlation of polymorphisms (r 2) at 500 nt. For the four neutral genes r 2 at 500 nt was 0.2 or less for teosinte samples, 0.3 or less for landrace samples, and 0.5 or less for inbreds (Table 2) consistent with previous reports (Remington et al., 2001; Tenaillon et al., 2001; Wright et al., 2005). For the candidate loci, in only two genes was the LD in teosintes above 0.2, AY107952, which is not considered a selected gene, and AY107821 with a high LD of 0.9. Analysis of AY107821 is complicated by having only four accessions successfully sequenced in the teosintes (Table 1), possibly yielding an artificially high LD. Many, but not all, of the genes judged significant for selection in the landrace versus teosinte comparison (domestication genes) have a substantial increase in the LD among the landrace accessions. An interesting gene in reference to LD is AY105850. Although this gene was determined to be an improvement gene by CS analysis of the extended sequence, the LD of this gene increases from 0.2 to 0.9 in comparing teosintes to the landraces. Visual examination of the pattern of polymorphisms in AY105850 (Fig. 3A) shows extensive haplotype structure in the landraces compared to the teosintes and distinctive patterns compared to another improvement gene AI737881 (Fig. 3B). This pattern of polymorphism may indicate some level of selection before the formation of the landraces. If the causal polymorphism in this gene was not a new mutation at the start of selection, but rather was low in frequency and distributed in a limited number of haplotypes, selection at domestication could give the observed pattern of retaining intermediate diversity, along with high LD in landraces (Innan and Kim, 2004). Overall, the CS analysis suggests that most of the selection in these genes occurred before the formation of the modern day landraces, and LD patterns are consistent with this conclusion.
Nucleotide polymorphisms for SNP and insertion/deletions at (A) AY105850 and (B) AI737881. Each horizontal rows indicates a line of maize inbreds, landraces, or teosintes, each vertical columns represent the nucleotide polymorphism position. Only polymorphic sites are indicated.
While the evidence points to a single domestication event of maize (Matsuoka et al., 2002) this is clearly not the case for crop improvement. Different inbred lines of maize have been derived from distinct germplasm pools at numerous locations around the world. Clearly additional theoretical and experimental research needs to be done to improve models for determining the genetic consequences of crop improvement and for testing for genes impacted by recent selection. Our current methods of analysis may be more effective in detecting selection at domestication than at crop improvement.
Conclusions
In summary, our results indicate that extended sequencing supported selection over genetic drift for a total of 15 of 27 selection screen candidates tested. Therefore, selection screens in maize are effective in identifying candidate genes involved in domestication and improvement. The added statistical support gained from extended sequencing permits the strongest selection candidates to be advanced into functional studies to investigate the causal sites within the genes and how selection on specific genes has altered plant phenotype.
With few exceptions, the area of reduced diversity for the genes judged as false positives ended shortly after the initially sequenced region (Fig. 1 and 2). Thus, it appears that if the length of the initial genome scan is increased from the ∼300 to 400 nt in Wright et al. (2005) and Yamasaki et al. (2005) to 600–800 nt the false-positive rate would decrease dramatically. The required size of the region sequenced for other species will be dependent on the diversity level of the progenitor and the strength of the domestication bottleneck. With recent advances in parallel sequencing technologies it may soon be possible to compare the gene space of numerous accessions with a species. This should permit a true genomic screen for artificial selection. However, the sequencing and assembling processes will need to be sufficiently powerful to make unambiguous assemblies of gene length units to limit the false-positive rates in candidate gene detection, or as in this study, initial candidates will need to be confirmed by extended gene sequencing.
Acknowledgments
This research was supported by National Science Foundation Plant Genome Award DBI0321467, by research funds provided by USDA-ARS (M.D.M.), and by the Japan Society for the Promotion of Science Postdoctoral Fellowship for Research Abroad in 2004 (M.Y.). We thank Kate Guill for technical assistance. Names of products are necessary to report factually on available data; however, neither the USDA nor any other participating institution guarantees or warrants the standard of the product, and the use of the name does not imply approval of the product to the exclusion of others that may also be suitable.
Footnotes
-
↵ * Corresponding author (mcmullenm{at}missouri.edu).
-
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.
-
- Received September 25, 2007.
- Accepted May 22, 2008.
- Copyright © 2008 Crop Science Society of America