Cereal crops, such as rice, wheat, and maize, are the world's most important sources of calories for humans, livestock feed for animals, and raw material for biofuel. Around 35% of human’s calorie intake comes from these crops (Ross-Ibarra et al., 2007). Production of these three major crops has increased greatly over the last 60 years, associated with improvement in farming technology, selection of higher-yielding varieties, the use of fertilizer, and sustainable intensification of agriculture. In the past ∼10,000 years, the major cereal crop varieties were domesticated and improved from their wild ancestors, driven by human selection, cultivation practices, and agricultural environments (Cowling et al., 2009; Gross et al., 2010; Olsen et al., 2013). During domestication, wild crop ancestors and following local landraces underwent strong selection to make the cultivated crops easy to culture and high yield, such as no shattering after maturation and increasing adaptation (Gross et al., 2010). Especially after adopting new technologies from a century ago, such as hybrid breeding, dwarf wheat and rice varieties, and genetic modification by transformation, the preliminary cultivated crop germplasms were collected and selected to render modern crops with high productivity and widespread adaptability (Baenziger et al., 2006). However, the Food and Agriculture Organization predicts that the global population will expand to over 9.2 billion in 2050 (http://www.fao.org/wsfs/forum2050/wsfs-background-documents/issues-briefs/en/), and agricultural production has to be increased by about 70 percent from the current levels to meet the increased food demand. Under the threat of limited arable land for cereal crop production by urbanization, land erosion, sea level rise, and pollution (Yu and Li 2021), sustainably enhancing cereal crop yield is critical to meet future demand.

To advance crop breeding, some essential clues come from understanding genetic mechanisms controlling their domestication (Fernie and Yan 2019). In the past three decades, investigating crop domestication, including where, when, and how it occurred, shed light on crop histories and mining favorable alleles from wild species to develop modern cultivars (Doebley et al., 2006). The domestication of major cereals, such as rice, maize, and wheat, began independently ∼10,000 years ago (Baenziger et al., 2006; Doebley et al., 2006; Fernie and Yan, 2019; Maccaferri et al., 2019; Smýkal et al., 2018; Fornasiero et al., 2022; Levy et al., 2022). Recent studies in crop comparative genomics revealed that Asian and African cultivated rice were domesticated ∼9,000 years ago in China and ∼3000 years ago in Africa, respectively (Fornasiero et al., 2022). The modern maize was domesticated from its wild ancestor, teosinte, ∼9,000‒10,000 years ago in southern Mexico (Doebley et al., 2006). For wheat, the domestication started ∼10,000 years ago in the Fertile Crescent (Maccaferri et al., 2019; Levy et al., 2022; Wang et al., 2023). After domestication, those preliminary crop varieties underwent improvement to pyramid beneficial mutations and recombinants in key genes and achieved high productivity and widespread adaptability (Doebley et al., 2006; Fernie and Yan 2019; Fornasiero et al., 2022; Levy et al., 2022). Therefore, crop domestication and improvement have reshaped the appearance and architecture of crop plants, converting from a low-productive wild species into a highly productive cultivated crop grown today.

Cereal crops underwent complex and independent selection in history (Baenziger et al., 2006; Doebley et al., 2006; Fernie and Yan 2019; Maccaferri et al., 2019; Guo et al., 2021; Fornasiero et al., 2022; Levy et al., 2022). However, a set of similar key traits were targeted by early communities, such as seed/fruit size, easy harvesting, widespread adaptability, etc (Hufford et al., 2012; Huang et al., 2012; Pankin and von Korff, 2017; Liang et al., 2021). Recent studies by genetic analysis of these traits have revealed that many domestication-related loci or genes play similar functions across species (Takeda et al., 2003; Studer et al., 2011; Lin et al., 2012; Dixon et al., 2018; Chen et al., 2022). For example, a gain-of-function allele of TEOSINTE BRANCHED1 (TB1) was identified to suppress tillering in domesticated maize through an increase in apical dominance compared to its wild ancestor teosinte (Studer et al., 2011). In rice and wheat, TB1 orthologs were found to function similarly to regulate lateral meristem initiation and tillering negatively (Takeda et al., 2003; Dixon et al., 2018). Another recently published study demonstrates that KRN2, which encodes a WD40 protein, negatively regulates grain number in rice and maize (Chen et al., 2022). The knock-out alleles of KRN2 by genome editing showed a further enhancement in grain yield in rice and maize (Chen et al., 2022). Therefore, the forward genetic studies in essential genes exploring, especially for these under convergent selection genes across the cereals, could help to understand the quantitative variation of important traits.

In the last century, the grain yield of cereal crops has risen steadily after adopting new technologies, such as hybrid breeding and high-yielding dwarf crop varieties (Fernie and Yan 2019; Liang et al., 2021). From the late nineteenth to middle twentieth centuries, plant breeders made many advances in breeding technologies, such as yield measurement at scale field trials, controlled crossings, hybrid breeding, utilization of ‘‘green revolution’’ genes, and pedigree-based estimates of breeding values, which started the Breeding 2.0 stage (Fernie and Yan 2019). Breeding 3.0 started 30 years ago, and several modern biotechnologies have been widely used in crop breeding, especially transgenic technologies and marker-assisted selection using molecular markers (Fernie and Yan 2019). We are now at the beginning of the Breeding 4.0 stage, driven by big biological data and the rapid progress of informatics technologies (Wallace et al., 2018; Fernie and Yan 2019). However, a full-chain organizational system has not yet been formed for germplasm resource utilization, gene mining, varieties development and industrial application. From the 1.0 to 4.0 stage of crop breeding, technological revolutions were the wheels that drove the breeding processes for the continuously increasing pace, accuracy, and precision (Wallace et al., 2018; Fernie and Yan 2019).

The artificial intelligence (AI) technology was suggested to facilitate crop breeding in the current 4.0 stage (Wallace et al., 2018; Fernie and Yan 2019). AI has gone through three stages with rapid development. The first stage was from the 1950s to the 1960s, when the concept of artificial intelligence was proposed. Machine translation with logical reasoning as its core mainly manifests in knowledge expression, such as propositional logic, predicate logic, and heuristic search algorithms. The second stage was in the 1970s and 1980s, when expert systems were proposed (Farina et al., 2024). The research and development of algorithms based on artificial neural networks are rapid. With the gradual improvement of semiconductor technology and computing hardware capabilities, artificial intelligence is gradually breaking through, and distributed networks have reduced the computational cost of artificial intelligence (Farina et al., 2024). The third stage is that since the end of the 20th century, especially since 2006, the era of cognitive intelligence has entered an era that emphasizes data and independent learning (Farina et al., 2024). Since genomes were sequenced, crop genomics has contributed significantly to plant breeding efforts toward novel varieties development, from understanding the genetic basis to systematic improvement for these important agronomic traits (Sun et al., 2022). Moreover, as a foundational tool, crop genomics initiates a data-driven era for 21st-century agriculture and plays a critical role in reshaping crop architecture for future needs (Sun et al., 2022). In this review, we summarized the most recent advances in biological insights for cereal crops from these data-driven studies and proposed the integration of multiple streams of data and insights into crop biology for an intelligence-driven crop breeding application in agriculture (Fig. 1).