OMICS- AND AI-ASSISTED BREEDING IN PULSE CROPS

Abstract

Pulses, including chickpea, pigeonpea, lentil, mungbean, blackgram, pea, cowpea and common bean, are an affordable source of dietary protein and enhance soil fertility through biological nitrogen fixation. The productivity of pulses remains low owing to drought, heat stress, diseases, insect pests, a narrow genetic base and long breeding cycles. Recent advances in the application of omics tools and artificial intelligence (AI) to pulse breeding are summarised herein. Reference genomes are now available for most major pulses, ranging in size from less than 0.6 Gb in mungbean and blackgram to approximately 13 Gb in faba bean. In addition, 3,366 chickpea accessions have been resequenced. Pan-genomes, single-nucleotide polymorphism (SNP) arrays, quantitative trait locus sequencing (QTL-seq) and genome-wide association studies (GWAS) have accelerated gene and marker discovery. Marker-assisted backcrossing has delivered Fusarium wilt-resistant chickpea varieties and drought-tolerant introgression lines in India. Transcriptomics, proteomics, metabolomics and phenomics provide complementary insights into gene function and plant performance under stress. Genomic selection (GS) studies in chickpea, pea, common bean and lentil have reported prediction accuracies of 0.14–0.63 for grain yield and up to 0.91 for seed weight. Machine learning (ML) and deep learning (DL) are increasingly applied to genomic prediction, image-based phenotyping, disease diagnosis and the integration of genomic, phenomic and environmental data. These models do not consistently outperform conventional linear models. Major challenges include small training populations, poor-quality phenotypic data, limited computational infrastructure and skilled personnel, and the need for explainable models. Integrating these tools with speed breeding, genome editing and envirotyping can shorten breeding cycles and accelerate the development of climate-resilient, nutritionally enhanced pulse varieties.