Eleusine coracana L. or finger millet is a traditional cereal grain that has been cultivated in arid and semiarid regions of Africa and Asia for thousands of years. This hardy crop is becoming increasingly popular because it can grow well under harsh environmental conditions such as drought and poor soil fertility, thereby serving as an important food source for areas that experience frequent famines1. The ability to survive on less water and nutrients resources and by transferring them more efficiently through its C4 photosynthetic pathway contributes to the adaptability of finger millet to different climates2.

Finger millet, often called a “super grain” is packed with essential nutrients. It serves as a great source of calcium, iron, & dietary fibre3. This helps a lot in promoting good health, especially for those at risk of malnutrition4. Notably, its calcium level is higher than that found in most cereals, benefiting bone health & helping prevent osteoporosis. Moreover, finger millet contains phenolic compounds and antioxidants, which improves its nutritional value & offers protective benefits against diseases like degeneration5. In the face of climate change & the challenges surrounding global food security, finger millet stands out as an excellent alternative to widely grown crops like rice & wheat. This crop can grow with very little input, making it a great choice for sustainable agriculture. This is especially true in smallholder farming systems6.

India plays a huge role in global millet production. It accounts for 44% of the total production in the world and 80% of Asia’s total finger millet production7. Millet cultivation spans roughly 32.5 million hectares in India, with finger millet occupying about 12.5% of that area, specifically, ragi millet is grown on 1.01 million hectares and produces around 1.67 million tons of yield8. Key producing states include Karnataka, Tamil Nadu, and Uttar Pradesh—Karnataka being the top producer at 1396.96 thousand tons while Uttar Pradesh contributes about 49.518 thousand tons9.

Finger millet production has great benefits, but it also faces many challenges. One of the biggest challenges hampering its production is blast disease, which is caused by the fungus Pyricularia grisea and is a serious devastating biotic stress faced by the crop. This disease brings about necrotic oval-shaped spots on the leaves. These spots look a lot like eyespots—they are tapered at both ends, wider in the middle, with greyish centres and dark-brown edges10, It can lead to early plant death with severe yield losses if not managed and addressed properly11,12. The pathogen can infect the crop at many growth stages which makes its impact even stronger. Yield losses may hit as high as 80% in the right conditions13. Warm weather, 25–28 °C temperature, with high humidity (80–90%), sets up the perfect atmosphere for fungi to thrive. Frequent rain or wet leaves for long periods further increase the chances of infection14. This is a big threat for areas that rely on finger millet for dietary food source & source of income.

Developing cultivars that resist blast disease is a very effective way to control this problem. Screening germplasm for resistance forms an important base in breeding programs. It helps in identification of resistant genotypes that can be used for breeding. This study focuses on evaluating a collection of finger millet germplasm for resistance to blast disease. Our goal is to find resistant genotypes and analyse the genetic structure of resistant germplasms using DNA fingerprinting by simple sequence repeats (SSR) markers. The SSRs or microsatellites markers are highly polymorphic which are most widely used due to their locus specificity, high abundance, multi-allelic property, co-dominance nature, high reproducibility and better stability. They are short, repetitive DNA sequences that vary in length among individuals, making them highly informative for genetic diversity studies, cultivar identification, and breeding programs15. By examining the resistance levels of different accessions, we hope to identify strong candidates for breeding programs aimed at improving blast resistance.

Materials and methods

Plant material

This study utilized and focused on a diverse group of 200 finger millet germplasm accessions. Alongside them, we included two known blast-resistant varieties: KMR204 & GE4449 and two susceptible control varieties, KMR301 and Udura mallige, to serve as control. These germplasm accessions were obtained from the Indian Institute of Millet Research (IIMR), Hyderabad.

Experimental site and design

The experiment took place at the Crop Research Farm, College of Agriculture at Banda University of Agriculture and Technology, Banda, Uttar Pradesh, India which is located at 25.5 °N, 80.3 °E and 113 AMSL. The weather data of experimental site during the cropping period is presented in Table 1. In this experiment, the diverse plant material evaluated in augmented block design during kharif season (from August 2023 to December 2023) and the experimental period lasted from 90 to 140 days. Four control varieties-two resistant & two susceptible, were planted after every 20 rows. This was done to make sure there was disease pressure throughout the experimental field. Each accession was planted in a net plot area of 0.6 square meters with 2 m row length, and spacing of 30 cm between the rows. Standard practices for finger millet cultivation were properly followed16. This included proper irrigation & fertilization, essential for optimal growth conditions. Importantly, no chemical fungicides were used throughout the experiment. This approach allowed for natural disease progression to occur without interference.

Table 1 Weather data during the cropping period.
Full size table

Pathogen identification

Diseased finger millet plants were gathered from areas where blast was prominent and showed clear blast symptoms, especially on leaves. We cut these leaves into small sections, about 1–2 cm long. Each section had blast lesions but also included a bit of healthy tissue at the end. Samples were sterilized with 70% alcohol separately. After sterilization, every piece was carefully placed onto oatmeal agar. This was done with sterilized needles & forceps to avoid contamination. The cultures were then incubated, and daily observations were made for the spore colony of Pyricularia grisea. Each individual spore colony was collected and moved to fresh oatmeal agar for purification & multiplication. Under a microscope, spores were examined for the typical blast conidiophores (Fig. 1).

Fig. 1
Fig. 1
Full size image

Observation of Pyricularia grisea (blast fungus) spores under microscope.

Disease screening

Leaf blast severity

Disease severity was evaluated each week after symptoms appeared. This followed a standardized 1–9 scale for leaf blast17 (Fig. 2). A score of 1 indicated there are either no lesions or just tiny brown specks, about the size of a pinhead (0.1–1.0 mm), impacting less than 1% of the leaf area. A score of 2 shows 1–5% of the area covered with lesions; 3 indicates 6–10% area affected, 4 indicates an area affected between 11 and 20%, while scores from 5 to 8 progressively indicate larger areas impacted: 21–30%, 31–40%, and up to 51–75%, with many leaves dead and a score of 9 means that over 75% of the area has blast lesions or complete leaf death. Accessions that display minimal symptoms, scoring between 1 and 3, are considered resistant. For accessions with moderate symptoms, scores from 3.1 to 5.0 categorize them as moderately resistant. If the disease scores are between 5.1 and 7.0, those accessions are classified as susceptible. Lastly, any accessions with scores from 7.1 to 9.0 fall into the highly susceptible category.

Fig. 2
Fig. 2
Full size image

Disease symptoms appeared on leaf of ragi plants under natural epiphytotic condition in experimental material at Crop Research Farm, BUAT, Banda. 1.0-2.0- highly resistant, 2.1-3.0- resistant, 3.1-5.0- moderately resistant, 5.1-7.0- susceptible and 7.1-9.0- highly susceptible.

Neck blast severity

While for neck blast severity, data collection was done following the initiation of panicle formation until maturity using 1–5 scale17 (Fig. 3). Here, a score of 1 indicates no lesions or just very small pinhead-sized ones on the neck region; a score of 2 shows lesions sized between 0.1 and 2.0 cm; a score of 3 indicates lesions measuring between 2.1 and 4.0 cm; a score of 4 for lesions that are between sizes of 4.1 and 6.0 cm and a score of 5 indicates lesions larger than 6 cm on the neck for neck blast disease. Accessions were classified based on these scores: scores from 1.0 to 2.0 are resistant; scores from 2.1 to 3.0 are moderately resistant; anything between 3.1 and 4.0 indicates susceptibility; while scores from 4.1 to 5.0 suggest high susceptibility.

Fig. 3
Fig. 3
Full size image

Symptoms of neck blast appeared on Ragi plants under natural epiphytotic condition in experimental material at Crop Research Farm, BUAT, Banda. 1.0-2.0- resistant, 2.1-3.0- moderately resistant, 3.1-4.0- susceptible, 4.1-5.0- highly susceptible.

Finger blast severity

For finger blast severity, observations were carried out after finger formation during the panicle development stage until maturity was reached using standardized 1 to 5 scale17 (Fig. 4). On this scale, a score of 1 means less than 2% of fingers are affected, while score 2 represents those with 2.1–10% affected fingers; a score of 3 indicates 10.1–20% affected; a score of 4 shows 20.1–30% affected; and 5 indicates over 30% fingers are affected by finger blast infection. Accessions were classified similarly as for neck blast categorization.

Fig. 4
Fig. 4
Full size image

Disease symptoms on fingers of Ragi plants under natural epiphytotic condition in experimental material at Crop Research Farm, BUAT, Banda. 1.0–2.0- resistant (1.0–10%), 2.1-3.0- moderately resistant (10.1–20%), 3.1-4.0- susceptible (20.1–30%) and 4.1-5.0- highly susceptible (> 30%).

Disease severity index

It was calculated according to formulae given by McKinney18.

$$DSI(\% )\, = \,\frac{{Sum\,of\,all\,disease\,ratings \times 100}}{{Total\,number\,of\,ratings\, \times \,Maximum\,disease\,grade}}$$

Area under disease progress curve. It was calculated using the following formulae given by Madden and coworkers19.

$$AUDPC\, = \,\frac{{\sum\limits_{{i = 1}}^{{n - 1}} {y_{i} + y_{i} + 1} }}{2}(t_{i} + 1 - t_{i} )$$

where “t” is the time of each reading, “y” is percent disease severity at each reading and “n” is the number of readings.

Agronomic traits

Morphological traits such as the number of days to 50% flowering (the period between sowing and when 50% of plants reach 50% anthesis in the main ear head) and days to maturity (the duration from planting until over 90% of plants show ear head browning before drying), plant height (determined in centimeters by the help of a meter scale from the plant’s base to the tip of the primary ear head), ear shape (fist type, compact type, semi compact type, open type and droop type), 1000 seed weight(g) and seed yield per plant (g) were observed and noted.

DNA fingerprinting

The experiment focused on characterizing resistant finger millet genotypes took place in the Molecular Plant Breeding Laboratory at the Department of Genetics and Plant Breeding, CoA, BUAT, Banda. A total of 20 SSR markers available in public domain were employed for this study, the list is mentioned in Table 2. For DNA extraction, young leaf tissues were collected for DNA fingerprinting from resistant and susceptible germplasm. DNA extraction followed a standard method known as cetyltrimethylammonium bromide (CTAB) method for DNA extraction10. After that, quantification was done with a Nanodrop spectrophotometer. To assess DNA quality, 0.8% agarose gel electrophoresis was used. For the genetic analysis, Simple Sequence Repeat (SSR) markers were employed. The selection included a total of 20 SSR markers specifically suited for finger millet, based on earlier research findings. In the PCR amplification process, a total reaction volume of 20 µL was created. This consisted of 10 µL of DreamTaq Green PCR Master Mix, 6 µL of nuclease-free water, 1 µL of working primer solution, & 1 µL of genomic DNA sample. The PCR process started with an initial denaturation step at 95 °C for 3 min followed by 35 cycles of denaturation (94 °C for 30 s), annealing at a specific temperature optimized for each primer pair for 30 s, & extension at 72 °C (1 min). Finally, there was an extension step held at 72 °C for a total of 7 min. The PCR products were separated on a 3% agarose gel stained with ethidium bromide. Visualization of the bands was done using a UV transilluminator Gel Documentation system. The SSR bands were then evaluated and noted as present (1) or absent (0) for each sample.

Table 2 List of SSR primers used in the research.
Full size table

Data analysis

R statistical package facilitated the statistical analysis. Analysis of variance (ANOVA) of augmented block design was conducted to highlight significant differences among the germplasm accessions. Furthermore, Pearson correlation coefficient analysis examined associations among the three types of blast disease with each other, & also between all morphological traits and each type of blast separately. Additionally, genetic variability parameters, estimates of heritability, and genetic advance for various traits were assessed as well. Molecular data analyses were carried out using both DARwin & PowerMarker software.

Results

Analysis of germplasm for blast resistance

From a total of 200 screened finger millet germplasm accessions for resistance, several showed promising results: IC0474832, IC0473539, IC0473580, IC0473822, IC0473823, IC0473864, IC0473880, IC0331685 & IC0331687 demonstrated combined resistance against all three levels of blast disease viz. leaf blast, neck blast, and finger blast under natural epiphytotic conditions (Table 3). For leaf blast specifically, nine accessions displayed strong resistance with consistently low disease scores and effective resistance mechanisms. In contrast, 91 accessions were moderately resistant, they showed some symptoms but nothing severe. 96 accessions were found to be susceptible with high disease scores; the remaining 4 were highly susceptible and showed severe leaf disease symptoms. For neck blast, a total of 162 accessions displayed resistance to blast disease, in contrast, 14 showed moderate resistance. 22 responded as susceptible and just 2 were highly susceptible according to mean neck blast severity. For finger blast, 155 were categorized as resistant while 20 were moderately resistant. Another 23 proved to be susceptible and 2 fell under the highly susceptible category based on the finger blast mean severity (Table 4).

Table 3 Disease response and AUDPC score of finger millet germplasm.
Full size table
Table 4 Categorization of germplasm according to leaf, neck and finger blast disease response.
Full size table

In the present study, the environmental data of kharif-2023 highlights the strong influence of climatic conditions on the pressure and spread of blast disease in finger millet (Table 4). Disease pressure refers to the intensity and likelihood of disease development, which in this case, was strongly driven by specific environmental factors like temperature, humidity, and rainfall patterns. It indicates that the temperatures between 27.3 °C (average minimum) to 33.3 °C (average maximum) created perfect conditions for leaf blast disease to thrive. Also, a relative humidity of 90.00% (avg. RH) was closely linked to higher incidence & severity of this disease. Long periods with high humidity were favourable for rapid spread of blast fungal growth and infection. Additionally, frequent rains averaging 48.6 mm (avg. RF) were related with an increase in disease occurrence. For neck & finger blast, low temperatures from 18.7 °C (average minimum) to 30.25 °C (average maximum) alongside moderate relative humidity of 65.30% played a role. Interestingly, rainfall did not appear to impact the disease progression for either neck or finger blast.

Analysis of variance

The analysis of variance (ANOVA) provided insights into the genetic variability among the traits under examination. The mean square values related to different blocks were not significant across all traits tested. This finding suggests that variability among blocks did not greatly impact the differences observed in these traits, indicating a consistent environmental influence throughout. Conversely, significant mean square values appeared when comparing genotypes to controls for all traits except for days until 50% flowering. Additionally, mean squares attributed to treatments & controls proved statistically significant for all traits at a 5% significance level.

Genetic variability parameters

Findings on genetic variability regarding all traits studies shows that the phenotypic coefficient of variation (PCV) consistently exceeds the genotypic coefficient of (GCV). The analysis of variability disclosed significant values for both GCV & PCV across multiple traits (Table 5). Notably, GCV values (greater than 20%) were found for 1000 seed weight (21.64%), seed yield per plant (56%), leaf blast (22.16%), neck blast (50.23%), finger blast (53.67%), &PC (53.23%). Meanwhile, moderate GCV values fell between 10 and 20% for days to 50% flowering (13.71%), days to maturity (10.25%), & plant height (15.28%). Similar patterns emerged for PCV as well; high values (> 20%) were noted in 1000 seed weight (23.83%), seed yield per plant (57.74%), leaf blast (22.56%), neck blast (55.02%), finger blast (57.42%), & AUDPC (53.42%). Moderate PCV values showed up in days to 50% flowering (14.3%), days to maturity (10.54%), & plant height (15.41%). Importantly, none of the values fell into the low category for both PCV and GCV, as per Deshmukh et al. (1986) classification.

Estimation of heritability and genetic advance as percent of mean

All characters studied showed high heritability, exceeding 60%, based on Robinson et al.20. classifications. This includes AUDPC with highest value of 99.27%, followed by plant height (98.32%), seed yield per plant (97.42%), leaf blast (96.48%), days to maturity (94.57%), days to 50% flowering (91.97%), finger blast (87.34%), neck blast (83.36%), and 1000 seed weight (82.4%). Furthermore, the GAM was also high across all evaluated traits, surpassing 20% according to Johnson et al.21. classifications. This includes seed yield per plant (116.04), AUDPC (109.41), finger blast (103.47), neck blast (94.62), leaf blast (44.9), 1000 seed weight (40.52), plant height (31.26), days to 50% flowering (27.13) and days to maturity (20.56) (Table 5).

Table 5 Range, general mean, GCV, PCV, heritability, genetic advance and genetic advance as per cent of mean for morphological characters and blast disease in finger millet genotypes.
Full size table

Correlation coefficient analysis

The estimates of correlation between blast disease and agronomic traits of finger millet is presented in Fig. 5. Leaf blast showed a notable, moderately weak positive correlation with both neck blast (r = 031**) & finger blast (r =0 0.31**). On the other hand, the correlation between neck & finger blast was very strong (r = 0.97**). Interestingly, leaf blast also displayed a significant negative correlation with days to 50% flowering (r = −0.16*), seed yield per plant (r = −0.26**), & plant height (r = −0.38**). There were also non-significant negative correlations noted with days to maturity (r = −0.13) & 1000 seed weight (r = −0.08). Neck blast exhibited a notable negative relationship with days to 50% flowering (r = −0.49**), days to maturity (r = −0.45**), plant height (r = −0.29**), & seed yield per plant (r = −0.27**). Still, it showed a non-significant negative correlation with 1000 seed weight (r = −0.10). Finger blast demonstrated a significant negative correlation with days to 50% flowering (r = −0.51**), days to maturity (r = −0.48**), plant height (r = −0.31**), & seed yield per plant (r = −0.26**). It also had a non-significant negative correlation with 1000 seed weight (r = −0.09).

Fig. 5
Fig. 5
Full size image

Diagrammatic representation of phenotypic correlation between various agronomic traits and disease severity in evaluated finger millet germplasm. (LB Mean leaf blast severity, NB Mean neck blast severity, FB Mean finger blast severity, DF Days to 50% flowering, DM Days to maturity, PH Plant height, SW 1000 seed weight, SY Seed yield per plant).

DNA fingerprinting

DNA fingerprinting involved characterization of eleven genotypes (Supplementary Figure S1 to S20). This included nine resistant & two susceptible genotypes. For the molecular characterization, twenty Simple Sequence Repeat (SSR) markers were used. Out of these, ten primers showed polymorphism among the finger millet genotypes. The polymorphic primers selected for further analysis included SSR2, SSR5, SSR7, SSR9, SSR10, SSR11, SSR14, and SSR17. Together, these ten primers amplified 26 alleles in total. The number of alleles ranged from a minimum of 2 (for SSR2, SSR5, SSR7, SSR9, and SSR17) to a maximum of 4 (specifically for SSR14), resulting in an average of 2.6 alleles per locus (Table 6).

Table 6 SSR markers displaying polymorphism among 11 finger millet germplasm lines.
Full size table

Discussion

Finger Millet is an important food in many parts of Asia and Africa but faces significant problems due to crop diseases caused by Pyricularia grisea. This disease poses a major threat to crop production and often causes large and significant losses hampering food security and livelihoods of those who rely on this crop. A clear understanding of the factors affecting the occurrence and intensity of millet blast is important to develop effective strategies to control and reduce the disease. As evident from the results, the present study investigates the genetic variation for blast resistance among the test genotypes, and role of environmental factors viz., temperature, relative humidity and rainfall on the spread of blast disease in finger millet, providing future breeding opportunities to increase blast disease resistance.

Major findings revealed that the accessions displayed a broad range of resistance levels. These levels ranged from highly susceptible to resistant. For instance, mean leaf blast severity went from 8.4, which indicates high susceptibility, to 1.7, showing resistance. Similarly, neck blast severity varied from 1 (resistant) to 4.4 (susceptible), while finger blast severity showed a range from 1 (resistant) to 4.6 (susceptible). This indicates significant genetic diversity among the accessions. Such variation creates a valuable genetic resource for breeding programs that aim to boost blast resistance and create new resistant varieties. These results suggest that disease severity was greatest at the leaf stage, with less progression to the neck and finger stages. Previous studies corroborate our findings, for instance, Babu et al.17. developed effective rating scales for all three blast types while proposing a screening method. Other researchers like Naik et al.22. , Patro et al.23. , Hemalatha et al.24. , Bal et al.6, Manyasa et al.25, Dida et al.26, Patro et al.27, Upamanya et al.28 also noted significant differences in blast resistance across various germplasm collections in both field and greenhouse conditions.

Warm temperatures with high humidity were best suited for leaf blast disease occurrence. On the other hand, neck and blast need cooler temperatures along with high to moderate humidity levels. These specific conditions promotes the spread of spores. Managing humidity in the canopy, crop rotation, early sowing practices can escape blast stress and lower the risk of infection. Extended periods of humidity after rainfall also aided in disease progression. According to Netam et al.29. , a minimum temperature of 21 °C and a maximum of 29 °C, combined with relative humidity levels between 70 and 81%, greatly contribute to the development of neck & finger blast disease, similar findings were obtained by Patro et al.30. Pal et al.31. found similar weather effects on rice blast.

The analysis showed significant variability among 200 genotypes across nine different traits, indicating substantial level of genetic diversity present. However, days to flowering revealed non-significant difference between genotypes and the control samples, which suggests either minimal genetic variation or a strong influence from environmental factors on this specific trait. Anuradha et al.32 conducted a study on finger millet genotypes and reported significant genetic diversity across multiple agronomic traits. Their findings highlighted the potential for resistance breeding programs. Upadhyay et al.33 investigated the genetic diversity in finger millet germplasm and observed considerable variability in traits such as plant height, grain yield, and days to maturity.

Research findings suggest that for all the traits examined, the phenotypic coefficient of variation (PCV) consistently surpassed the genotypic coefficient of variation (GCV). This implies that environmental factors have a greater impact than genetics on variability. High GCV and PCV values for traits such as blast severity and seed yield point towards significant potential for genetic improvement. Traits like days to flowering and plant height showed moderate values; they indicate some variation but still offer chances for selection. The lack of low values in both PCV and GCV across traits suggests there is substantial genetic diversity and opportunities for breeding advancements. These findings align with Ramana et al.34. reported moderate to high GCV values for most of the yield and yield-contributing traits, the highest GCV and PCV values included grain yield per plant, plant height, and the number of productive tillers per plant. Angadi et al.35 also reported that the PCV values were generally higher than the GCV values for most of the traits assessed, including blast resistance.

All traits displayed high heritability, suggesting a robust genetic influence. Genetic advance as a percentage of the mean (GAM) was also significantly high across all traits, especially in seed yield per plant & AUDPC. This indicates to great potential for enhancement selective breeding. Ravikumar & Seetharam36 had similar results, noting high heritability in disease resistance and other traits. Anuradha et al.37. found high heritability along with high GAM across various characteristics like seed yield, days to flowering, and multiple disease traits (including leaf blast, neck blast, & finger blast).

Regarding findings on correlation, a weak positive correlation between leaf blast and both neck & finger blast was found which suggests that these traits are less related. In contrast, neck blast and finger blast showed a very strong positive correlation which indicates they are closely linked. Interestingly, leaf blast negatively correlated with days to flowering. This suggests that later-maturing varieties could avoid peak disease pressure. Similar outcomes were reported by Thakur et al.38. , Babu et al.17. , among others. Seed yield showed a negative relationship with leaf blast. This is primarily because the disease adversely affects plant health & grain development. Our results are consistent with those of Mgonja et al.39. A similar negative correlation was observed in plant height. Taller plants may be able to avoid or resist the disease, likely due to specific microclimatic factors. Comparable results were noted by Nagaraja et al.40. , Babu et al.17. , and Wekesa et al.41. for correlation of blast with plant height. Neck & finger blast also had negative correlations with flowering time, maturity, height, and yield. These strong relationships emphasize their similar effects on these traits. Moreover, the correlation between all three types of blast & finger shape indicated no correlation since the disease was seen across all panicle shapes—compact, semi-compact, droopy, open, and fist. Babu et al.17. supported these findings and conclusions well.

Conclusion

The genotypes IC0474832, IC0473539, IC0473580, IC0473822, IC0473823, IC0473864, IC0473880, IC0331685, and IC0331687 showed resistance to all three types of blast disease. These genotypes hold potential value as cultivars for breeding programs. Leaf blast occurs most often in warm conditions. Specifically, it thrives at an average minimum temperature of 27.3 °C & an average maximum of 33.3 °C. High humidity between 90 and 95% along with moderate rainfall greatly contributes to leaf blast development. On the other hand, cooler temperatures & lower humidity tend to favor neck & finger blast. A moderate positive correlation was found between leaf blast with both neck and finger blast types. However, neck & finger blast types have a strong positive correlation with one another. When looking at all three kinds of blast disease, they show a negative correlation with the traits we studied. This suggests that as the disease gets worse, the performance in those traits decreases. Every trait had high heritability and a significant genetic advance percentage of the mean, which means selective breeding for these traits is likely to work well.

In summary, the findings highlight the significant potential of breeding programs that utilize resistant genotypes as cultivars. They also emphasize how environmental factors affect development. Furthermore, there are promising opportunities for genetic improvement through selection.