Melon (Cucumis melo L.) is a globally important horticultural crop [1]. China is the world’s top melon producer, with its cultivation area and output accounting for 36.23% and 49.09% of the global total, respectively [2]. However, the intensification of global climate change and the scarcity of water resources have rendered drought stress one of the primary abiotic stresses impacting the growth, development, yield formation, and fruit quality of horticultural crops [3,4]. As a representative arid region, Xinjiang has scarce precipitation and insufficient irrigation water resources. Drought and water scarcity often lead to a substantial decline in melon production [5]. Consequently, screening and identifying drought-resistant varieties are crucial measures for the development of stress-resistant crops and the guarantee of agricultural production [6].

Seed germination is the most sensitive stage for plants in response to water deficit. The drought tolerance at this stage directly determines the field seedling emergence rate and is closely associated with the subsequent crop yield [7]. Drought is the primary stress factor restricting seed germination and plant growth. Inadequate water supply notably impedes seed metabolism and hinders the germination process [8]. As stress intensifies, the seed germination rate, germination potential, germination index, and vigor index all exhibit a significant decline [9].

In response to drought stress, plants activate a series of drought-tolerance response mechanisms. The elevation of malondialdehyde (MDA) content acts as a crucial indicator of oxidative stress induced by drought. As a product of membrane lipid peroxidation, it can be used to assess the degree of cell membrane damage and plant drought tolerance [10,11]. In the context of osmotic regulation, proline (Pro) and soluble protein (SP) act as crucial osmotic regulatory substances. Their contents gradually increase with the intensification of drought stress, which helps maintain cell osmotic pressure and mitigate cellular damage [12,13]. Plants rely on their antioxidant system to counteract drought-induced oxidative damage. This system primarily consists of superoxide dismutase (SOD), peroxidase (POD), ascorbate peroxidase (APX), and catalase (CAT) [14]. Drought leads to a substantial accumulation of reactive oxygen species (ROS) and impairs the activity of various enzymes in plants [15]. Meanwhile, the activities of enzymes such as CAT, APX, SOD, and POD typically increase significantly to scavenge excessive reactive oxygen species (ROS) and mitigate their harmful effects on cell membranes, chloroplasts, and enzyme structures, thereby enhancing drought tolerance [4,16].

Polyethylene glycol (PEG-6000) has been extensively used in the screening of drought-resistant crop germplasms due to its stable stress-inducing effect and high controllability [17]. Based on this method, previous researchers have conducted drought-resistant germplasm screening in multiple crop species. For example, Ahmed et al. [18] identified the drought tolerance of 80 different wheat genotypes and selected wheat lines with strong drought tolerance, such as NR-499, NARC-2009, and Pakistan-2013. Similar methods have been applied to the screening of drought-resistant lines of Cucumis sativus L. Reference [19] and Zea mays L. [20]. Current research on melon drought resistance primarily focuses on physiological responses and drought-resistance gene mining at the seedling stage, and there are still notable gaps in the systematic drought-resistance screening of melon germplasm resources [21,22]. Although Xinjiang has abundant region-specific indigenous melon germplasm resources, a large-scale systematic identification of drought resistance during the germination stage for these local germplasms has not been carried out to date. Existing studies on the drought resistance evaluation of melon germplasm at the germination stage mostly use a single phenotypic trait, such as germination rate, as the evaluation basis and lack a multi-dimensional comprehensive evaluation system that integrates germination growth indicators, osmotic adjustment substances, and antioxidant enzyme activities [2]. Furthermore, few studies have integrated relevant physiological and biochemical indicators at the germination stage to construct drought resistance prediction models, making it difficult to achieve rapid drought resistance grading for large batches of germplasm. In view of these limitations, this study selected 60 native melon germplasm accessions from Xinjiang to screen for local germplasm resources with high drought tolerance and integrated germination morphological and physiological indicators to construct a drought resistance prediction model, aiming to fill the gap in the rapid drought resistance evaluation system for Xinjiang melon germplasm at the germination stage.

Traditional evaluations of plant drought tolerance are predominantly based on a single germination trait, such as germination rate (GR). Although this approach can reveal phenotypic differences among different germplasms, it cannot fully characterize the complex physiological and metabolic regulatory networks in plants under drought stress. Evaluation based on a single indicator is also highly vulnerable to environmental fluctuations, which often leads to biased assessment results [2,23]. In this study, the multi-index membership function method was used to integrate multiple types of indicators, including germination traits, osmotic adjustment substances, and antioxidant enzyme activities, into a comprehensive evaluation system. The comprehensive evaluation value D was calculated based on the coefficient of variation for each indicator, which effectively reduces the deviation caused by single-indicator evaluation and enables a more objective assessment of the overall drought tolerance of the tested germplasms. On this basis, core indicators were further screened through stepwise regression analysis, and a corresponding prediction model was constructed. This method simplifies the measurement process while guaranteeing evaluation accuracy and provides technical support for rapid drought tolerance screening of large-scale germplasm resources.

In this study, 60 local melon germplasm resources from Xinjiang were used as experimental materials. Polyethylene glycol 6000 (PEG-6000) was employed to simulate drought stress, and drought tolerance at the germination stage was evaluated to screen for highly drought-tolerant germplasms. A rapid drought tolerance screening model was established, and key physiological and biochemical indicators for drought tolerance prediction were selected. This research provides a scientific basis and a technical approach for drought tolerance breeding, elite parent selection, and large-scale, accurate, and rapid drought tolerance screening of melon germplasms in the arid regions of Xinjiang.

A concentration of 20% PEG was identified as the screening concentration, which was consistent with the stress concentration adopted by Wu et al. [5] for drought-resistant cultivar screening of Xinjiang Hami melon. The germination rate, germination potential and radicle length of cucurbit seeds gradually decreased with increasing PEG concentration, which was in line with previous studies on Cucumis sativus L. [24], Glycine max L. Merr. [25] Brassica napus L. [26], Zea mays L. Reference [27] and Cucurbitaceae [28]. It is worth noting that the radicle length of some melon germplasms, such as L36 in this study, increased first and then decreased with the increasing PEG concentration, which was consistent with the conclusion of Taminždić et al. [29] in the drought stress test of cucumbers with different genotypes. Similar phenomena were also observed in Zea mays L. Reference [30] and Lens culinaris Medik. [31]. This suggests that under moderate stress, the plant defense system may be more effectively stimulated, ultimately enabling the plant to grow better under drought stress conditions [32]. When plants perceive water deficit, this low-concentration-induced root elongation phenomenon may also occur: plants activate the root osmotic regulation system and actively maintain the division capacity of root tip cells, thereby promoting water absorption from deeper soil layers [33]. Melon seed germination was completely inhibited under 30% PEG stress, yet some halophytes and woody species can still maintain a certain level of germination activity under the same intensity of osmotic stress [34,35]. In contrast, melon, as a typical horticultural crop, is more sensitive to water deficit during seed germination.

Cluster analysis demonstrated that drought-tolerant germplasms exhibited higher osmotic adjustment capabilities and antioxidant enzyme activities. As documented in prior research, plants accumulate osmolytes (e.g., SP, Pro) and enhance the activities of antioxidant enzymes such as SOD, CAT, APX, and POD in response to drought stress [36,37], which is consistent with the results of this study. Cucumber [23], watermelon [38] and other cucurbit crops [28] also showed similar physiological response characteristics, all of which reduced oxidative damage caused by drought by increasing osmotic adjustment and various antioxidant enzyme activities. The results of the correlation analysis indicate that antioxidant indicators including Pro, SOD and CAT show significant positive correlations under drought stress. However, these results only reflect correlational relationships at the numerical level and cannot confirm the existence of direct regulatory causal associations between these indicators. It is hypothesized that proline and the two antioxidant enzymes mitigate oxidative damage and osmotic stress induced by drought via synergistic responses. The plant drought response is a complex biological process regulated by multiple pathways, and plant drought resistance relies on the synergistic action of multiple functional substances. An increase in the activity of a single enzyme may not lead to a significant improvement in overall drought tolerance [39]. The aforementioned physiological indicators merely reflect response characteristics under drought stress, while the upstream and downstream molecular regulatory mechanisms underlying these responses still require further validation through subsequent experimental investigation. In this investigation, the activities of POD and SOD in certain drought-sensitive melon germplasms, such as L35 and L53, decreased after 7 d of stress. This phenomenon can be ascribed to the imbalance of the antioxidant system and excessive accumulation of reactive oxygen species in cucurbit crops under severe or long-term drought conditions [40]. The bidirectional alteration of enzyme activity resulting from this genotypic difference is not an anomaly. Some studies have exposed three different melon genotypes to PEG stress, finding that SOD and POD activities in sensitive varieties continued to decline, while those in drought-tolerant varieties increased steadily [41]. Antioxidant enzyme activities in pumpkin [42], melon [43], and watermelon [38] showed varying degrees of increase or decrease under drought stress. In contrast, antioxidant enzymes in most gramineous crops typically exhibit a unidirectional increase under drought conditions, with only a few instances of decrease [44]. This phenomenon implies the diversity and genotype-specific nature of drought tolerance and antioxidant mechanisms in cucurbit crops.

Some studies have found that after drought stress, drought-tolerant melon germplasms accumulate less MDA, which is consistent with the results of this study [1]. Owing to the higher efficiency of the antioxidant and osmotic regulation systems in drought-tolerant varieties, they generally exhibit lower MDA content [11]. Their content can be used to assess the degree of cell membrane damage and plant drought tolerance [10,11]. Similar studies have shown that MDA content in drought-resistant wheat [45] and melon cucumber [39] increased less. Under osmotic stress induced by 20% PEG, the increase in MDA content in germplasm with high drought tolerance was significantly lower than that in drought-sensitive germplasm. As MDA is the final product of membrane lipid peroxidation, these results indicate that highly drought-tolerant germplasm may possess a more stable membrane lipid composition or stronger membrane repair capacity under osmotic stress [46]. After 7 d of 20% PEG stress, the SP and Pro contents of melon increased, which was consistent with the response trends of Citrullus lanatus, Glycine max (L.) Merr. [13], and Solanum lycopersicum L. Reference [12] under drought stress. In our study, the increment of proline content under 20% polyethylene glycol stress was greater than that of other biochemical indicators. This may be because plants synthesize abundant proline to maintain normal osmotic adjustment under water deficit conditions [35]. Previous studies have demonstrated that plants can alleviate membrane damage by accumulating proline [40], while exogenous proline can boost the activities of antioxidant enzymes and mitigate oxidative injury [47]. These findings are consistent with the stress responses observed in wheat [48] and celery [49]: proline plays a vital role in reactive oxygen species detoxification, preservation of membrane integrity and stabilization of multiple enzyme activities. Accordingly, elevated proline levels can strengthen drought tolerance in plants [50]. Under osmotic stress, the average increase in proline content was significantly higher than that of soluble protein. Specifically, proline content in some germplasm materials reached as high as 388%. This result indicates that melon tends to adopt the proline biosynthesis pathway as the main osmotic adjustment strategy, rather than relying on the accumulation of soluble protein [5,46].

In this study, the germination rate and root growth of 27 melon germplasm accessions were severely inhibited under 20% polyethylene glycol (PEG) stress. Based on such extreme germination phenotypes, these accessions were directly classified as extremely drought-sensitive genotypes. This evaluation method has precedent in research on other crops. For instance, Yadav et al. [51] identified the tomato cultivar Srijana as drought-sensitive since it failed to germinate completely under PEG stress. Mannan and Begum [52] adopted a similar approach to assess germination phenotypes of 97 maize germplasm accessions and categorized those with poor performance as drought-sensitive. A multi-index comprehensive evaluation method was applied to the remaining 33 accessions available for biochemical determination. Moreover, a single indicator cannot fully reflect drought tolerance, whereas comprehensive evaluation with multiple indicators can markedly improve the reliability of evaluation results [2,23]. The 27 melon germplasm resources with low germination rates and radicle lengths of less than 0.3 cm were classified as drought-sensitive types. This threshold was determined based on measured data from melon germination experiments under 20% PEG-induced drought stress. Under this stress condition, only a small number of seeds can germinate, and the longest radicle among germinated seeds was only 0.3 cm. Seedling radicles of this length do not have sufficient fresh weight, making it impossible to obtain complete data for all physiological and biochemical indicators. To standardize the sampling protocol and ensure the comparability of experimental data, 0.3 cm was selected as the critical classification threshold.

In our research, 33 melon germplasms were divided into three categories via clustering analysis of 10 indexes using membership function values. Similar methods have been applied to drought tolerance evaluation of 13 watermelons [38] and 58 melons [1]. Additionally, four key drought tolerance indexes (Pro, RL, MDA, SOD) for the melon germination stage were screened out through a drought tolerance prediction model constructed via stepwise regression. CAT and SOD showed collinearity and were preferentially eliminated during screening. The remaining unselected indexes had weak correlations with melon drought tolerance at the germination stage, failing to effectively improve model fitting accuracy and prediction ability, so they were excluded from the model. Compared with the prediction model constructed by Ren et al. [2] using data from 15 melon genotypes at the seedling stage, this study has a larger sample size and focuses on the germination stage—the most drought-sensitive stage for melons and a key stage determining seedling growth in production. Consequently, the model is more closely aligned with actual production scenarios in arid regions and has higher reliability and application value. In this study, stepwise regression was used to screen the core indicators of drought resistance, which could eliminate the multicollinearity of indicators and adapt to multi-index joint modeling. However, this method is limited by the dataset and lacks generalization, making it necessary to expand the germplasm samples to verify the stability of the results. It should be emphasized that the established model is only applicable to the preliminary screening of drought-tolerant melon germplasm during the seed germination stage, and its application still has certain limitations. In field cultivation, melons grown in Xinjiang face the risk of drought stress throughout the entire growth period.

In this study, 27 accessions were excluded from the modeling dataset because their poor germination under 20% PEG stress prevented the collection of sufficient radicle tissue for physiological assays. Given that model construction requires complete empirically measured datasets, forced inclusion of these incomplete records would inevitably introduce statistical bias. Conversely, modeling based exclusively on complete-case data enhances the reliability and robustness of the predictions. Accordingly, the established model is specifically applicable to the preliminary large-scale screening of germplasms with normal germination capacity. It is important to note that the model is confined to the Xinjiang local melon germplasm resources evaluated in this study and is limited to the germination stage; extrapolation to other germplasm populations or growth stages is not recommended. The results should be interpreted as a reference for subsequent refined identification rather than as definitive conclusions. For the 27 sensitive accessions that failed to pass the germination screening, their drought susceptibility has already been phenotypically confirmed. Should further physiological characterization be required, alternative assessments using lower concentrations of PEG-6000 at the seedling stage may be considered. Furthermore, in this study, both model construction and fitting performance evaluation were performed using the same dataset, and no independent external validation set was established. Although the adjusted coefficient of determination is as high as 0.989, this modeling strategy still carries a risk of overfitting. Specifically, while the model shows excellent adaptability to the germplasm data collected in this experiment, its prediction accuracy will likely decrease when applied to melon germplasms from other batches or other sources. Accordingly, future research needs to expand the independent sample dataset for external validation to further improve the generalizability of the model. Among the 60 Xinjiang local melon germplasm populations examined in this experiment, RL, Pro, MDA and SOD were the most informative indexes for drought resistance evaluation only at the germination stage. These indicators cannot be regarded as universal drought screening markers for all melon materials until further verification using independent germplasm populations from other geographical regions is completed.

Stepwise regression analysis was used to screen four core drought resistance indices, namely Pro, RL, MDA and SOD. Among these indices, root length is the only morphological index. During the germination stage, excellent root elongation ability can expand the water absorption range of seedlings and reduce the damage caused by drought stress, representing an intuitive morphological response of melon to early drought resistance [39]. Under field conditions, there may be a potential association between root length at the germination stage and drought resistance at the adult plant stage. Melon seedlings with outstanding root elongation ability under drought stress during germination have stronger early water uptake capacity and improved water absorption ability, which helps them accumulate growth advantages for subsequent development. This root-related advantage may persist into the adult stage, promoting the formation of deep root systems in the field and enhancing the long-term soil drought tolerance of plants. However, complex environmental factors in the field, such as soil structure, can change the growth trajectory of roots [53]. Therefore, root length at the germination stage and the drought resistance of adult plants may not be completely synchronized, and the relationship between the two remains unclear at present.

In this study, 60 local melon germplasms sourced from Xinjiang were subjected to PEG-6000 simulated drought stress to systematically screen and comprehensively evaluate their drought tolerance during the germination stage. A 20% polyethylene glycol (PEG) concentration was within the semi-lethal range for the tested germplasms. In conjunction with comparisons of the coefficients of variation for germination rate across various stress levels, 20% PEG was selected as the optimal concentration for drought tolerance screening. Drought stress significantly inhibited seed germination and radicle elongation. Germplasms with high drought tolerance exhibited superior osmotic adjustment capacities, heightened antioxidant enzyme activities, and less severe cell membrane damage. Correlation analysis indicated that osmotic adjustment (Pro) collaborated with antioxidant enzymes (SOD, CAT) to alleviate drought stress in melon germplasms under the 20% PEG treatment. The 60 germplasms were classified into four groups: highly tolerant, moderately tolerant, low-tolerant, and drought-sensitive. The 14 highly drought-tolerant germplasms identified can be employed as elite parental materials for drought tolerance breeding. A stepwise regression model, D = −0.309 + 0.053 × Pro + 0.319 × RL + 0.469 × MDA + 0.137 × SOD, was established to rapidly predict drought tolerance at the germination stage and improve the efficiency of large-scale drought tolerance screening in melon germplasms. In addition, in this study, model construction and fitting performance evaluation were performed using the same dataset, and no independent external validation set was established. Therefore, future research needs to expand the independent sample dataset for external verification to further improve the generalizability of the model. PEG-induced osmotic stress does not fully represent field drought conditions, only the germination stage was evaluated, and molecular validation of the selected physiological indicators was not performed.