Lekgari Lekgari
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Lekgari Lekgari has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Motlalepula Pholo-Tait
Faculty of Research and Graduate Studies, Research Centre for Bioeconomy, Botswana University of Agriculture and Natural Resources, Private Bag 0027, Gaborone 00267, Botswana
2 shared publications - Samodimo Ngwako
Department of Crop and Soil Sciences, Botswana University of Agriculture and Natural Resources, Private 0027, Gaborone 00267, Botswana
2 shared publications - Aobakwe Boisy Tshegofatso
Department of Crop and Soil Sciences, Botswana University of Agriculture and Natural Resources, Gaborone 00267, Botswana
1 shared publication - Bamphitlhi Tiroesele
Botswana University of Agriculture and Natural Resources
1 shared publication Chenamani Amos
1 shared publication- Galalea Gillian Gaonosi
Department of Crop and Soil Sciences, Botswana University of Agriculture and Natural Resources, Gaborone 00267, Botswana
1 shared publication
Latest Publications
- Open Access
- 20 October 2026
- Published by Elsevier in SSRN Electronic Journal
Classification of Botswana Sorghum Seed Varieties Using Deep Convolutional Neural Networks
Correct identification of pure sorghum seed varieties is important for breeding and seed production, but the traditional methods employed require specialized skills not available to all stakeholders in the seed value chain, especially in Botswana. Deep learning using convolutional neural networks (CNN) is a new approach to seed classification which is potentially more accurate and accessible than the traditional methods. Therefore, seven CNN models were assessed for possible deployment in a mobile computing platform as a base model for accurate and automatic classification of three sorghum seed varieties from Botswana. The effect of topological depth and network design on classification performance was investigated using the following CNN models of topological depths ranging from seven to 389 layers: AlexNet, VGG16, Xception, MobileNetV2, InceptionV3, DenseNet121, and NASNetMobile. In addition, model size was taken into consideration as a factor influencing portability. Results showed that the models with lower topological depths performed better than the ones with higher depths, with test losses and accuracies, respectively, ranging from 0.17 to 0.77 and 0.67 to 0.94. Correspondingly, Precision, Recall and F1-scores ranged between 0.67 and 0.94 across all models evaluated. However, the networks of lower topology produced very large models, while the ones with higher topological depths produced smaller sized models. In conclusion, the DenseNet121 model is recommended for optimization for deployment as a base model for a mobile sorghum seed classification system owing to its good prediction performance (accuracy and F1 score of 0.8), and its lean size of only 54.9MB and 10,146,891 parameters.
- 20 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessClassification of Botswana Sorghum Seed Varieties Using Deep Convolutional Neural Networks
Correct identification of pure sorghum seed varieties is important for breeding and seed production, but the traditional methods employed require specialized skills not available to all stakeholders in the seed value chain, especially in Botswana. Deep learning using convolutional neural networks (CNN) is a new approach to seed classification which is potentially more accurate and accessible than the traditional methods. Therefore, seven CNN models were assessed for possible deployment in a mobile computing platform as a base model for accurate and automatic classification of three sorghum seed varieties from Botswana. The effect of topological depth and network design on classification performance was investigated using the following CNN models of topological depths ranging from seven to 389 layers: AlexNet, VGG16, Xception, MobileNetV2, InceptionV3, DenseNet121, and NASNetMobile. In addition, model size was taken into consideration as a factor influencing portability. Results showed that the models with lower topological depths performed better than the ones with higher depths, with test losses and accuracies, respectively, ranging from 0.17 to 0.77 and 0.67 to 0.94. Correspondingly, Precision, Recall and F1-scores ranged between 0.67 and 0.94 across all models evaluated. However, the networks of lower topology produced very large models, while the ones with higher topological depths produced smaller sized models. In conclusion, the DenseNet121 model is recommended for optimization for deployment as a base model for a mobile sorghum seed classification system owing to its good prediction performance (accuracy and F1 score of 0.8), and its lean size of only 54.9MB and 10,146,891 parameters.
- Open Access
- 5 August 2025
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 26 (15) , 7543
Estimating Genetic Variability and Heritability of Morpho-Agronomic Traits of M5 Cowpea (Vigna unguiculata (L.) Walp) Mutant Lines
Induced mutation plays an integral part in plant breeding as it introduces new variability among the population. A study was conducted in cowpeas [Vigna unguiculata (L.) Walp] to assess the yield divergence, heritability, genetic advance, and correlation among the M5 Tswana cowpea mutants. The experiment utilized seven genotypes under rainfed and supplementary irrigation during the 2022/23 and 2023/24 cropping seasons. The mutant lines demonstrated significant variations in days to 50% emergence (DE) and days to 50% flowering (DF). Tswana emerged earlier (5–7 days) and flowered in 21–54 days across the two seasons, compared to some of the mutant lines. The yield and yield components varied among some mutant lines and the control. Most importantly, mutants outperformed the Tswana control for some of these traits, indicating potential for genetic improvement. An analysis of genetic parameters revealed minimal environmental influences on some of the observed traits (GH, PN, GY), while others showed little environmental impact. Variation in heritability (H2) and genetic advance (GA%) between the two seasons limited the contribution of genotypic effects in the expression of the studied traits. Correlation analysis revealed strong and significant positive associations between DE and GH, as well as between DF and PW. Most traits, except DF and PW, were positively correlated with grain yield (GY), although the correlations were not significantly different. Cluster analysis grouped the genotypes into four distinct clusters. Principal component analysis (PCA) revealed the superiority of mutant lines (Tswana-300Gy-214, Tswana-400Gy mutant lines, and Tswana-500Gy-31) in their association with improved GY, pod weight (PW), 100-seed weight (100-SW), and seed number per pod (SN/P). Interestingly, the Tswana control formed a separate cluster and diverged from the mutants in PCA, suggesting that induced mutagenesis effectively targeted genes controlling the traits considered in this study.
- 5 August 2025
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 26 (15) , 7543
Open AccessEstimating Genetic Variability and Heritability of Morpho-Agronomic Traits of M5 Cowpea (Vigna unguiculata (L.) Walp) Mutant Lines
Induced mutation plays an integral part in plant breeding as it introduces new variability among the population. A study was conducted in cowpeas [Vigna unguiculata (L.) Walp] to assess the yield divergence, heritability, genetic advance, and correlation among the M5 Tswana cowpea mutants. The experiment utilized seven genotypes under rainfed and supplementary irrigation during the 2022/23 and 2023/24 cropping seasons. The mutant lines demonstrated significant variations in days to 50% emergence (DE) and days to 50% flowering (DF). Tswana emerged earlier (5–7 days) and flowered in 21–54 days across the two seasons, compared to some of the mutant lines. The yield and yield components varied among some mutant lines and the control. Most importantly, mutants outperformed the Tswana control for some of these traits, indicating potential for genetic improvement. An analysis of genetic parameters revealed minimal environmental influences on some of the observed traits (GH, PN, GY), while others showed little environmental impact. Variation in heritability (H2) and genetic advance (GA%) between the two seasons limited the contribution of genotypic effects in the expression of the studied traits. Correlation analysis revealed strong and significant positive associations between DE and GH, as well as between DF and PW. Most traits, except DF and PW, were positively correlated with grain yield (GY), although the correlations were not significantly different. Cluster analysis grouped the genotypes into four distinct clusters. Principal component analysis (PCA) revealed the superiority of mutant lines (Tswana-300Gy-214, Tswana-400Gy mutant lines, and Tswana-500Gy-31) in their association with improved GY, pod weight (PW), 100-seed weight (100-SW), and seed number per pod (SN/P). Interestingly, the Tswana control formed a separate cluster and diverged from the mutants in PCA, suggesting that induced mutagenesis effectively targeted genes controlling the traits considered in this study.
- Open Access
- 6 May 2025
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 26 (9) , 4406
Phenotypic Screening and Marker-Assisted Validation of Sources of Aphis craccivora Koch Resistance in Cowpea (Vigna unguiculata L.)
Aphis craccivora significantly affects cowpea (Vigna unguiculata L.) production, leading to yield reductions. Management strategies encompass physical barriers and biological and chemical methods, which can be costly and detrimental to the environment. Host-plant resistance offers a more sustainable alternative. This study evaluated cowpea genotypes in a screenhouse experiment. Tswana and B261-B were resistant, while B301, B259, and ER7 showed a tolerance phenotype. Tswana exhibited a low aphid population and minimal plant damage, probably due to suppression of reproduction and fecundity. Conversely, IT97K-556-6, SARI-21KTA-6, SARC 1-57-2, B013-F, B339, and Blackeye were susceptible to aphids, as shown by high aphid populations and dense sooty molds. Severe damage to plant vigor may be linked to direct aphid feeding and reduced photosynthesis efficiency. SNP1_0912 and CP 171F/172R markers confirmed aphid resistance in Tswana and ER7 as well as in the IT97k-556-6 and SARI-21KTA-6 controls. The inverse susceptible phenotype in the control group suggests that the markers may not function properly due to negative interactions between quantitative trait loci (QTL) and environmental factors. This could also indicate the presence of different aphid biotypes that severely damage Western African breeding lines. This study offers essential insights for breeding aphid-resistant cowpea varieties. Future efforts will involve sequencing Tswana to identify more resistance sources and create novel markers.
- 6 May 2025
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 26 (9) , 4406
Open AccessPhenotypic Screening and Marker-Assisted Validation of Sources of Aphis craccivora Koch Resistance in Cowpea (Vigna unguiculata L.)
Aphis craccivora significantly affects cowpea (Vigna unguiculata L.) production, leading to yield reductions. Management strategies encompass physical barriers and biological and chemical methods, which can be costly and detrimental to the environment. Host-plant resistance offers a more sustainable alternative. This study evaluated cowpea genotypes in a screenhouse experiment. Tswana and B261-B were resistant, while B301, B259, and ER7 showed a tolerance phenotype. Tswana exhibited a low aphid population and minimal plant damage, probably due to suppression of reproduction and fecundity. Conversely, IT97K-556-6, SARI-21KTA-6, SARC 1-57-2, B013-F, B339, and Blackeye were susceptible to aphids, as shown by high aphid populations and dense sooty molds. Severe damage to plant vigor may be linked to direct aphid feeding and reduced photosynthesis efficiency. SNP1_0912 and CP 171F/172R markers confirmed aphid resistance in Tswana and ER7 as well as in the IT97k-556-6 and SARI-21KTA-6 controls. The inverse susceptible phenotype in the control group suggests that the markers may not function properly due to negative interactions between quantitative trait loci (QTL) and environmental factors. This could also indicate the presence of different aphid biotypes that severely damage Western African breeding lines. This study offers essential insights for breeding aphid-resistant cowpea varieties. Future efforts will involve sequencing Tswana to identify more resistance sources and create novel markers.