Using machine learning for image-based analysis of sweetpotato root sensory attributes.

The sweetpotato breeding process involves assessing different phenotypic traits, such as the sensory attributes, to decide which varieties to progress to the next stage during the breeding cycle. Sensory attributes like appearance, taste, colour and mealiness are important for consumer acceptability and adoption of new varieties. Therefore, measuring these sensory attributes is critical to inform the selection of varieties during breeding. Current methods using a trained human panel enable screening of different sweetpotato sensory attributes. Despite this, such methods are costly and time-consuming, leading to low throughput, which remains the biggest challenge for breeders. In this paper, we describe an approach to apply machine learning techniques with image-based analysis to predict flesh-colour and mealiness sweetpotato sensory attributes. The developed models can be used as high-throughput methods to augment existing approaches for the evaluation of flesh-colour and mealiness for different sweetpotato varieties. The work involved capturing images of boiled sweetpotato cross-sections using the DigiEye imaging system, data pre-processing for background elimination and feature extraction to develop machine learning models to predict the flesh-colour and mealiness sensory attributes of different sweetpotato varieties. For flesh-colour the trained Linear Regression and Random Forest Regression models attained R2 values of 0.92 and 0.87, respectively, against the ground truth values given by a human sensory panel. In contrast, the Random Forest Regressor and Gradient Boosting model attained R2 values of 0.85 and 0.80, respectively, for the prediction of mealiness. The performance of the models matched the desirable R2 threshold of 0.80 for acceptable comparability to the human sensory panel showing that this approach can be used for the prediction of these attributes with high accuracy. The machine learning models were deployed and tested by the sweetpotato breeding team at the International Potato Center in Uganda. This solution can automate and increase throughput for analysing flesh-colour and mealiness sweetpotato sensory attributes. Using machine learning tools for analysis can inform and quicken the selection of promising varieties that can be progressed for participatory evaluation during breeding cycles and potentially lead to increased chances of adoption of the varieties by consumers.

Connecting data for consumer preferences, food quality, and breeding in support of market-oriented breeding of root, tuber, and banana crops.

The 5-year project 'Breeding roots, tubers and banana products for end user preferences' (RTBfoods) focused on collecting consumers' preferences on 12 food products to guide breeding programmes. It involved multidisciplinary teams from Africa, Latin America, and Europe. Diverse data types were generated on preferred qualities of users (farmers, family and entrepreneurial processors, traders or retailers, and consumers). Country-based target product profiles were produced with a comprehensive market analysis, disaggregating gender's role and preferences, providing prioritised lists of traits for the development of new plant varieties. We describe the approach taken to create, in the roots, tubers, and banana breeding databases, a centralised and meaningful open access to sensory information on food products and genotypes. Biochemical, instrumental textural, and sensory analysis data are then directly connected to the specific plant record while user survey data, bearing personal information, were analysed, anonymised, and uploaded in a repository. Names and descriptions of food quality traits were added into the Crop Ontology for labelling data in the databases, along with the various methods of measurement used by the project. The development and application of standard operating procedures, data templates, and adapted trait ontologies improved the data quality and its format, enabling the linking of these to the plant material studied when uploaded in the breeding databases or in repositories. Some modifications to the database model were necessary to accommodate the food sensory traits and sensory panel trials. © 2023 The Authors. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

Targeting market segment needs with public-good crop breeding investments: A case study with potato and sweetpotato focused on poverty alleviation, nutrition and gender.

Crop breeding programs have often focused on the release of new varieties that target yield improvement to achieve food security and reduce poverty. While continued investments in this objective are justified, there is a need for breeding programs to be increasingly more demand-driven and responsive to the changing customer preferences and population dynamics. This paper analyses the responsiveness of global potato and sweetpotato breeding programs pursued by the International Potato Center (CIP) and its partners to three major development indicators: poverty, malnutrition and gender. The study followed a seed product market segmentation blueprint developed by the Excellence in Breeding platform (EiB) to identify, describe, and estimate the sizes of the market segments at subregional levels. We then estimated the potential poverty and nutrition impacts of investments in the respective market segments. Further, we employed the G+ tools involving multidisciplinary workshops to evaluate the gender-responsiveness of the breeding programs. Our analysis reveals that future investments in breeding programs will achieve greater impacts by developing varieties for market segments and pipelines that have more poor rural people, high stunting rates among children, anemia prevalence among women of reproductive age, and where there is high vitamin A deficiency. In addition, breeding strategies that reduce gender inequality and enhance appropriate change of gender roles (hence gender transformative) are also required.

Development of diagnostic SNP markers for quality assurance and control in sweetpotato [Ipomoea batatas (L.) Lam.] breeding programs.

Quality assurance and control (QA/QC) is an essential element of a breeding program's optimization efforts towards increased genetic gains. Due to auto-hexaploid genome complexity, a low-cost marker platform for routine QA/QC in sweetpotato breeding programs is still unavailable. We used 662 parents of the International Potato Center (CIP)'s global breeding program spanning Peru, Uganda, Mozambique and Ghana, to develop a low-density highly informative single nucleotide polymorphism (SNP) marker set to be deployed for routine QA/QC. Segregation of the selected 30 SNPs (two SNPs per base chromosome) in a recombined breeding population was evaluated using 282 progeny from some of the parents above. The progeny were replicated from in-vitro, screenhouse and field, and the selected SNP-set was confirmed to identify relatively similar mislabeling error rates as a high density SNP-set of 10,159 markers. Six additional trait-specific markers were added to the selected SNP set from previous quantitative trait loci mapping studies. The 36-SNP set will be deployed for QA/QC in breeding pipelines and in fingerprinting of advanced clones or released varieties to monitor genetic gains in famers' fields. The study also enabled evaluation of CIP's global breeding population structure and the effect of some of the most devastating stresses like sweetpotato virus disease on genetic variation management. These results will inform future deployment of genomic selection in sweetpotato.

Gene-based single nucleotide polymorphism markers for genetic and association mapping in common bean.

BACKGROUND: In common bean, expressed sequence tags (ESTs) are an underestimated source of gene-based markers such as insertion-deletions (Indels) or single-nucleotide polymorphisms (SNPs). However, due to the nature of these conserved sequences, detection of markers is difficult and portrays low levels of polymorphism. Therefore, development of intron-spanning EST-SNP markers can be a valuable resource for genetic experiments such as genetic mapping and association studies. RESULTS: In this study, a total of 313 new gene-based markers were developed at target genes. Intronic variation was deeply explored in order to capture more polymorphism. Introns were putatively identified after comparing the common bean ESTs with the soybean genome, and the primers were designed over intron-flanking regions. The intronic regions were evaluated for parental polymorphisms using the single strand conformational polymorphism (SSCP) technique and Sequenom MassARRAY system. A total of 53 new marker loci were placed on an integrated molecular map in the DOR364 × G19833 recombinant inbred line (RIL) population. The new linkage map was used to build a consensus map, merging the linkage maps of the BAT93 × JALO EEP558 and DOR364 × BAT477 populations. A total of 1,060 markers were mapped, with a total map length of 2,041 cM across 11 linkage groups. As a second application of the generated resource, a diversity panel with 93 genotypes was evaluated with 173 SNP markers using the MassARRAY-platform and KASPar technology. These results were coupled with previous SSR evaluations and drought tolerance assays carried out on the same individuals. This agglomerative dataset was examined, in order to discover marker-trait associations, using general linear model (GLM) and mixed linear model (MLM). Some significant associations with yield components were identified, and were consistent with previous findings. CONCLUSIONS: In short, this study illustrates the power of intron-based markers for linkage and association mapping in common bean. The utility of these markers is discussed in relation with the usefulness of microsatellites, the molecular markers by excellence in this crop.