Progress 05/01/12 to 04/30/17
Outputs
Target Audience:Across the extent of this project, our group systematically interacted with research working in plant scientists interested in applying genomic selection tools in their breeding programs and also those interested in developing and implementing genomic prediction tools in pragmatic breeding programs, especially in plants. These types of interactions occurred during scientific meetings, invitations that the PIs and/or graduate student working in the project received, to visit specific private companies or other public research institutions to give presentations or participate of panel discussion. Some of the specific venues for such interactions includes invited talks at Dow AgroSciences, ICRISAT 4th International Workshop on Next Generation Genomics and Integrated Breeding for Crop Improvement, a University of Minnesota Special Seminar, the Plant Genomics Congress, the Corn Breeding Research meeting as well as poster presentations at multiple Maize Genetics Conferences and the Crop Science Society of America annual meeting. In addition to that, both the PI and co-PI have received over the years numerous visitors from around the country interested in learning about the work developed in the project and visit the fields and the field testing sites where the different materials involved in this project were developed and evaluated. Additionally, a number of undergraduate research assistants have been trained through this project. Among the most relevant ones, Amanda Drinsinger, a graduate from the University of Wisconsin, Platteville, worked as a limited term employee as part of this project. Amanda is pursuing an advanced degree in the field of plant breeding and her experience working as part of this project offered her the opportunity to gain valuable hands on experience in the field that she is now directly applying to her own research. Also, Wilson Craine, who worked as a student hourly in this project, is currently a graduate student at Washington State University pursuing a PhD in plant science, with an emphasis in breeding, specially looking to improve Camelina (a potentially promising oilseed crop) for early season cold tolerance to widen its adaptation to more diverse production environments. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The focus of the initial part of this project was to develop the appropriate germplasm and genomic resources needed. The second part of this project involved a significant amount of field evaluation and selection. Seven undergraduate students participated of different aspects of this research during the length of this project. The experience gained through this involved has motivated three of them (Wilson Craine, Amanda Drinsinger, and Rachel Perry) to pursue advanced degrees in different aspects of plant science. This grant also supported a graduate student who has responsibilities to develop activities related to the simulation work addressing the relative efficiency individual vs multi-trait selection and also combining results from the GS and phenotypic based selection for the comparison of selection efficiency. A visiting postdoctoral fellow (Dr. Renato Rodrigues Silva), who was sponsored by the Brazilian government, contributed to the analysis of the effect of selection on diversity (G3 publication). How have the results been disseminated to communities of interest?In addition to numerous oral and poster presentations, a manuscript entitled "Selection for Silage Yield and Composition Did Not Affect Genomic Diversity Within the Wisconsin Quality Synthetic Maize Population" was published in 2015 in the journal G3 describing some of the results from the initial evaluation of the WQS population. Two other manuscripts are in preparation describing simulation work that compares selection for relative efficiency of selecting for individual traits compared to multivariate analysis and the other publication which will describe the comparison of the completed cycles of GS and phenotypic based selection. Phenotypic data for this final comparison was collected just this summer due to the delays caused by adverse climatic conditions in some of the scheduled evaluation seasons. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
The major impact of this project includes the confirmation that relatively simple genomic selection models provide sufficiently accurate prediction ability for most traits in plant breeding programs. Also, results from this project suggest that the ability to utilize historical information to inform decision making in ongoing breeding programs is limited due to the interaction of genotypes and environments, specially across years. Also, the efficiency of selecting traits individually vs combining them into index selection is depends on genetic correlations of the traits under selection. This project, initially, required developing of a number of experimental tools. During the first year of the project, a set of 438 S2 lines from cycle 0 (C0) to C4 of the WQS maize selection were genotyped using the Illumina 55K SNP DNA chip. A set of 15,610 SNP informative markers were used in combination with historical performance data generated by the program during the process of selection to define the most appropriate genomic selection (GS) model for each of the traits of interest which included forage yield, dry matter content and biomass compositional traits. A number of selection models were tested in this population. A ridge regression best linear unbiased prediction (RR-BLUP) approach provided the best prediction results and therefore it was the one used for subsequent selection. For the first two rounds of selection a set of 215 S1 families from WQS C5 and then a set of 192 S1 families from WQS C6, respectively, were genotyped using the same Illumina platform as for the training population and the best 20 lines were selected from each selection cycle utilizing the prediction model previously developed (testing population) in each case. In every case, a parallel set of 20 randomly selected lines was also advanced through the selection protocol as a negative control. This effort also maintained a S1 phenotypic based selection program running in parallel for comparison purposes. In general, moderate to high family-mean heritabilities were observed for the different traits within cycles of selection. On the other hand, variation across cycles and years accounted for, most of the realized additive relationship matrix, and therefore the ability to utilize information across generations was significantly more challenging. This evidence points to the idea that the interaction of genotype and year is a very relevant factor that can hinder the ability to exploit historical information for the implementation of GS in practical breeding programs. Another important finding from this work points to the idea that although overall variability (measured as gene diversity) has been maintained at a reasonable level across the genome in general, specific regions presented significant reduction in diversity. Those specific regions are likely the genomic location of genes or regulators controlling the traits under selection. The ability to associate those regions of low diversity with quantitative trait loci controlling the traits was limited in the WQS likely due to the relatively small population size (low power) and the innate population structure present in this population. The G3 manuscript published in 2015 summarizes some of these findings. A component of this project also focused on developing simulation analysis based on the experimental structure of WQS, genotypic information collected as part of this project and historical phenotypic information gathered as part of the selection process of the WQS population to assess the relative efficiency of individual trait vs index-based selection through the application of whole-genome molecular maker information. The primary observations from this work suggest that the relative efficiency of selecting for individual traits compared to multivariate analysis is highly dependent on the genetic covariance between traits under selection and such covariance are, in turn, dependent on the effect of selection on allelic frequency changes. Observations point to the idea that the relative advantage of one selection approach over the other depends not only on the genetic architecture of the traits under selection but also on the proportion of shared genetic control among those traits. A manuscript summarizing these findings is currently in preparation. The phenotypic evaluation of families from the phenotypic-based selection of WQS that was proposed as part of this project has been challenging overall. Only one of the two locations planned in 2013 could be planted due to excessive rain in the spring season. The evaluation was repeated again in 2014 in order to obtain a second year of data for decision making. This delay the phenotypic selection component of this project one year. In 2014, our first attempt was made to recombine the top 20 lines identified during the first cycle of GS selection. Germination on the portion of the nursery where these lines were planted was very poor due to suboptimal weather conditions immediately following planting. Given the importance of maintaining allelic balance during recombination, we decided to repeat the recombination step the following year. The set was planted again in the summer and pollinations were successfully completed. The set of 20 randomly selected lines was simultaneously advanced at each selection stage. Recombination of the individuals identified for the second cycled of selection was completed in 2015 and shelfed in 2016. The phenotypic evaluation of the resulting hybrids for the head-to-head comparison, therefore, took place in 2017. Results from this evaluation will be summarized in a publication currently in preparation.
Publications
- Type:
Other
Status:
Published
Year Published:
2014
Citation:
Genomic Selection The Maze of Maize International Innovation Research Media Ltd.
|
Progress 05/01/14 to 04/30/15
Outputs
Target Audience:During the May 2014 to April 2015 reporting period, we have continued our interactions with researchers working in the field of application of genomic selection tools in plant breeding breeding programs. This was mostly achieved by participating on scientific meetings including oral presentations describing how the level of genetic relationship among individuals affects prediction accuracy for genomic selection in plant breeding. The specific venues for these presentations included an invited talk by one of the PIs (Dr. Lorenz) at Dow AgroSciences with headquarters in Indianapolis, IN on October 28, 2014 and another one at the Plant Genomics Congress that took place in St. Louis, MO on September 11-12, 2014. In addition to that, Amanda Drinsinger, a recent graduate from the University of Wisconsin, Platteville, worked as a limited term (LTE) employee supported by this project. Amanda is interested in pursuing an advanced degree in the field of plant breeding and this LTE position (duration from October 2014 to May 2015) provided the opportunity for her to gain hands on experience on the management and logistics related to the application of a genomic selection approaches in a plant breeding program. She participated in activities such as nursery harvesting and seed and sample preparation and processing. Changes/Problems:Suboptimal weather conditions during the spring of 2014 affected both, our recombination nursery and one of our two testing sites for the phenotypically selected populations. Both activities had to be repeated during the summer of 2015, which represents a delay in the generation of appropriate germplasm resources and selection cycles. What opportunities for training and professional development has the project provided?The initial part of this project focused on developing the appropriate germplasm and genomic resources. As such, much of the efforts involved sample identification, greenhouse planting for tissue collection, and DNA extraction and subsequent genotyping. An undergraduate student research assistant was trained by one of our laboratory research specialist during the first year of this project. Last year, a recent graduate in the field of biology with an interest in plant breeding worked as a limited term employee. She was more involved in the field evaluation, sampling processing and logistics. Given her interest in pursuing an advanced degree in the area of plant breeding, this was a very worthwhile opportunity to acquire hands-on experience on the details involved in a breeding program handling phenotypic and genotypic information. How have the results been disseminated to communities of interest?A manuscript entitled "Selection for Silage Yield and Composition Did Not Affect Genomic Diversity Within the Wisconsin Quality Synthetic Maize Population" was published earlier this year in the journal G3 describing some of the results from the initial evaluation of the WQS population. Additionally, oral presentations utilizing portions of the data were done. Venues for these presentations included an invited talk by one of the PIs (Dr. Lorenz) at Dow AgroSciences with headquarters in Indianapolis, IN on October 28, 2014 and another one at the Plant Genomics Congress that took place in St. Louis, MO on September 11-12, 2014. What do you plan to do during the next reporting period to accomplish the goals?Recombined populations for both the most recent genotypically and phenotypicaclly selected populations that are being completed this summer will be selfed and then genotyped during spring 2016 to complete one more cycle of selection. The graduate student (Mr. Joseph Gage) working on this project continues to develop activities related to addressing objective 1 and portions of objective 3 of our proposal utilizing results from the initial cycle of selection and simulation work. A postdoctoral fellow (Dr. Renato Rodrigues Silva), who was sponsored by the Brazilian government, has conducted a significant amount of work related to objectives 2 and 4 of this project also combining results from the initial cycles of selection and simulations. We plan to complete two separate publications describing these two efforts before our next reporting cycle.
Impacts
What was accomplished under these goals?
The practical focus of this project has involved the simultaneous development and advancement of specific populations selected using genomic selection (GS) as well as phenotypic tools. During the first year of the project, we genotyped a set of 438 S2 lines from the WQS maize population cycle 0 (C0) to C4 using the 55K SNP chip developed by Illumina. This generated a useful set of 17,719 SNP markers that was combined with performance data collected over the entire course of the selection program. This set represents our training population and was used to develop genomic selection models for yield, dry matter content and cell wall compositional characteristics. The number of markers available allowed appropriate coverage of the genome based on the level of linkage disequilibrium decay observed. During initial stages of the project, we experimented with several different selection models. The ridge regression best linear unbiased prediction (RR-BLUP) model was selected to be used in all subsequent analysis of the data collected in this project. For the first round of genomic selection, a set of 215 S1 families from WQS C5 was genotyped using the Illumina platform used for the training population. From these we identified the best 20 lines utilizing the prediction model previously developed (testing population). During the spring of 2014, the second round of GS was performed. In this second round 192 S1 families from WQS C6 were genotypically evaluated and phenotypes predicted using the model previously developed. The top 20 lines were identified and a first attempt was made to recombine them during the summer 2014. Unfortunately, germination for this portion of the nursery was very poor due to suboptimal weather conditions immediately following planting, therefore the recombination was not completed. The set was planted again this summer and pollinations were just successfully completed. A corresponding set of 20 randomly selected lines is simultaneously being advanced through the pipeline at each selection stage as a negative control. Also in parallel, S1 based phenotypic selection activities continue underway to develop the appropriate materials for genotypic vs. phenotypic based selection comparisons. The initial phenotypic evaluation of S1 families from the phenotypic-based selection of WQS C5 was conducted during the summer of 2013 as planned, however excessive rain in the spring of 2014 severely affected one of our two locations, and therefore the evaluation had to be repeated in 2014. Selections were performed and recombination of selected individuals for this set is also being conducted this summer. Earlier this year we completed publication of the initial findings from this project. Overall, family mean heritabilities for the traits evaluated (yield and compositional characteristics) within cycles were moderate to high (from 0.32 to 0.82). However, the proportion of the variation accounted for by the realized additive relationship matrix (i.e.: transmittable variation) across cycles and years was substantially lower (0.09 to 0.42). This observation suggests that genotype-by-year interaction is an extremely influential factor for breeding programs that utilize information generated across seasons. As part of this initial evaluation we assessed the overall level of variability, measured as genetic diversity. Despite the severe selection pressure, variability has been maintained overall and observed localized reductions have not occurred uniformly across the genome. A few large regions of the genome have reach near fixed, while others have experience virtually no changes in their diversity levels. In addition to the continuous development of appropriate germplasm resources to test pragmatic questions, we continue to deploy different combination of simulation analyses based on the demographic structure of the WQS population complemented by genotypic and phenotypic data collected from the initial cycles of the this population to assess different aspects of the application of GS in breeding. The primary topics currently under study include: 1) the relative efficiency of individual trait vs. index-based selection for the application of whole-genome molecular maker information for selection; 2) the identification of appropriate training and testing populations as it relates to the longevity of selection models for complex traits.
Publications
- Type:
Journal Articles
Status:
Accepted
Year Published:
2015
Citation:
Lorenz, A.J., T.M. Beissinger, R. Rodrigues-Silva, Natalia de Leon (2015) Selection for Silage Yield and Composition Did Not Affect Genomic Diversity Within the Wisconsin Quality Synthetic Maize Population. G3 5:541-549 doi: 10.1534/g3.114.015263.
- Type:
Journal Articles
Status:
Accepted
Year Published:
2015
Citation:
Lorenz, A.J., and K.P. Smith. 2015. Adding genetically distant individuals to training populations reduces genomic prediction accuracy in barley. Crop Sci. doi: 10.2135/cropsci2014.12.0827.
|
Progress 05/01/13 to 04/30/14
Outputs
Target Audience: During the May 2013 to April 2014 reporting period, we have been able to reach researchers interested in the applicability of genomic selection into phenotypically based breeding programs. These interactions have been primarily done through participation on scientific meetings such as invited talks related to the topic of optimization of genomic prediction for public plant breeding programs at venues such as the Corn Breeding Research meeting (Chicago, IL Dec 8 and 9), ICRISAT 4th International Workshop on Next Generation Genomics and Integrated Breeding for Crop Improvement. (Hyderabad, India, Feb 19-21) and an invited presentation at University of Minnesota Special Seminar (St. Paul, MN, May 8). An outreach article was generated by International Innovation – Research Media Ltd. published in May 2014 entitled Genomic Selection – The Maze of Maize, which specifically highlighted the work being done in this project. In addition to that, Ben Kaeppler, an undergraduate research assistant participated of this project during July and August 2013. He participated in activities such as sample preparation, tissue collection and DNA extraction, which provided a rich opportunity for the training of this student in the field and laboratory. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided? The initial stages of this project involved a substantial amount of sample identification, greenhouse planting and tissue collection for DNA extraction and subsequent genotyping. An undergraduate research assistant was trained by one of our laboratory research specialist on those tasks. This was an important training opportunity for this undergraduate student. How have the results been disseminated to communities of interest? Initial results have been presented to several audiences primarily in oral format. Some of the venues where this has been presented include Corn Breeding Research meeting (Chicago, IL Dec 8 and 9), ICRISAT 4th International Workshop on Next Generation Genomics and Integrated Breeding for Crop Improvement. (Hyderabad, India, Feb 19-21) and an invited presentation at University of Minnesota Special Seminar (St. Paul, MN, May 8). What do you plan to do during the next reporting period to accomplish the goals? The recombined populations being generated this summer will be selfed and then genotyped during fall/winter 2014 to complete one more cycle of GS in this population. At the same time, the phenotypic-based selection portion of the project continues to advance despite the challenging weather conditions experienced during spring 2013. A graduate student (Mr. Joseph Gage) joined our program on 2013. He is responsible for work related to objective 1 and portions of objective 3. In addition to that, a postdoctoral fellow (Dr. Renato Rodrigues Silva), sponsored by the Brazilian government and is being part of our group since 2013 is mostly focused on addressing issues related to objectives 2 and 4 of this project. We are in the process of finalizing our first publication for submission later this summer or early fall which will describe the overall structure of the WQS population and the effect of selection on the genomic architecture of this population. The manuscript will also discuss the utility of this population for the identification of genomic signatures caused by selection.
Impacts
What was accomplished under these goals?
The focus of the first years of our project has been to develop the experimental aspects of the projects, which are the foundation to address the goals proposed. During the first year of the project, a set of 438 S2 lines from the WQS maize population cycle 0 (C0) to C4 were genotyped using the Illumina 55K SNP chip. We used set of 15,610 SNP markers in combination with performance data collected over the course of the selection program to develop genomic selection models for each trait (yield, dry matter content and cell wall compositional characteristics). Based on the level of linkage disequilibrium decay observed, the number and distribution of markers generated here appears to be appropriate. Several different selection models were tested, and a ridge regression best linear unbiased prediction (RR-BLUP) model was finally used in the analysis. During the first round of GS, a set of 215 S1 families from WQS C5 were genotyped using the 55K Illumina platform used for the training population and the best 20 lines were selected utilizing the prediction model previously developed (testing population). During the spring of 2014, the second round of GS is being performed. For this 192 S1 families from WQS C6 were evaluated genotypically and the top 20 selected lines are being recombined during summer 2014 for further advancement. A parallel set of 20 randomly selected lines is also being advanced through the pipeline at each selection stage as a negative control. S1 based phenotypic selection activities are also underway in parallel. From the initial cycles of WQS we observed moderate to high (from 0.32 to 0.82) family-mean heritabilities for individual traits within cycles. The proportion of variation across cycles and years accounted for by the realized additive relationship matrix, however, was substantially reduced (0.09 to 0.42). This observation supports the idea that genotype-by-year interaction is an important factor to be considered for breeding programs that exploit historical information. Additionally, the overall level of variability, measured as gene diversity, has been maintained at a reasonable level across all loci; however, gene diversity reductions have not occurred uniformly across the genome. Some large genomic regions have become nearly fixed, while diversity has been maintained at others. A manuscript summarizing these findings is currently in preparation. A simulation analysis based on the structure and genotypic and phenotypic data collected from the initial cycles of the WQS population was conducted to assess the relative efficiency of individual trait vs index-based selection through the application of whole-genome molecular maker information. Initial results suggest that the relative efficiency of selecting for individual traits compared to multivariate analysis is highly dependent on the genetic covariance between traits under selection. The first phenotypic evaluation of S1 families from the phenotypic-based selection of WQS C5 was conducted during 2013 as planned, however only one of the two locations planned could be planted due to excessive rain in the spring season. The evaluation is being repeated again in 2014 to have an additional location of data. This population will continue to be advanced using genomic and phenotypic strategies in order to assess long-term implications these different selection methods in the improvement of complex traits in diverse populations. Overall, tools and resources generated in this project are expected to assist maize breeders currently practicing phenotypic recurrent selection transition to genomic recurrent selection as the cost of genotyping continues to decline.
Publications
|
Progress 05/01/12 to 04/30/13
Outputs
Target Audience: During this reporting period, this project has been able to reach groups of plant breeders interested in issues related to the applicability of genomic selection tools into phenotypically based breeding programs. Such outreach activity has been primarily possible through interactions at scientific meetings, specially the Maize Genetics Conference that took place in St Charles, IL on March 14 to 17, 2013. A poster was presented at that conference describing the initial evaluations of the Wisconsin Quality Synthetic silage maize breeding populations and the discoveries related to genomic structure and architecture of this population that affect its feasibility for genomic selection approaches. Also, during the process of collecting genotypic data for this project, an undergraduate student research assistant participated of sample preparation, tissue collection and DNA extraction, which provided a nice opportunity for the training of this student in the laboratory. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided? The initial stages of this project involved a substantial amount of sample identification, greenhouse planting and tissue collection for DNA extraction and subsequent genotyping. An undergraduate research assistant was trained by one of our laboratory research specialist on those tasks. This was an important training opportunity for this undergraduate student. How have the results been disseminated to communities of interest? Initial results from the analysis of the structure and constitution of the WQS population and initial steps on the associative analysis of genotypes and phenotypes involved in this project was presented in poster format at the Maize Genetics Conference in St Charles, IL on March 2013. What do you plan to do during the next reporting period to accomplish the goals? The recombined populations generated during winter nursery 2012-13 have been planted in our summer nursery for selfing. Genotyping of those populations will be done during the winter season to complete an additional cycle of genomic selection. At the same time, the phenotypic-based selection portion of the project continues to advance as planned. A graduate student (Mr. Joseph Gage) joined our program on June 28th 2013 and will have major responsibilities related to the continuation of objective 1 and aspects of objective 3. In addition to that, a postdoctoral fellow (Dr. Renato Rodrigues Silva), sponsored by the Brazilian government, also started on May 1st 2013 and will be mostly focused on addressing issues related to objectives 2 and 4 of this project. We are in the process of finalizing our first publication for submission later this summer or early fall which will describe the overall structure of the WQS population and the effect of selection on the genomic architecture of this population. The manuscript will also discuss the utility of this population for the identification of genomic signatures caused by selection.
Impacts
What was accomplished under these goals?
During the first year, the main focus was to develop the experimental aspects of this project, primarily related to specific aims 1 and 2 described above. For that, a set of 438 S2 lines from the Wisconsin Quality Synthetic (WQS) maize population including lines from cycle 0 (C0) to C4 were genotyped using the Illumina 55K SNP chip technology. Marker data was combined with phenotypic data available as part of the University of Wisconsin Silage Breeding program to develop genomic selection models for forage yield and relevant compositional traits using ridge regression best linear unbiased prediction. The specific model of selection was determined after studies of cross validation conducted using the available data. A set of 215 S1 lines from WQS C5 were also genotyped using the same genotyping platform and the best 20 lines based on a combination of yield and compositional characteristics were selected utilizing the prediction model just developed (testing population). The set of 20 selected lines as well as a randomly selected set of 20 lines were sent to our 2012-13 winter nursery for recombination and generation of the first GS cycle. These experimental tools developed during the first year of this project represent the basis that will allow addressing the main objectives of this research.
Publications
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2013
Citation:
Lorenz, A.J., N. de Leon (2013) Optimal resource allocation for a maize genomic recurrent selection program. 55th Maize Genetics Conference. St Charles, IL, March 14-17
|
|