Cotton genotypes selection through artificial neural networks.

Area of Science:

• Agricultural Science
• Genetics
• Computational Intelligence

Background:

• Traditional breeding programs rely on statistical analysis for genotype selection.
• Computational intelligence, specifically artificial neural networks (ANNs), is underutilized in cotton genetic improvement.
• Improving cotton fiber quality is crucial for the textile industry.

Purpose of the Study:

• To explore the application of artificial neural networks as auxiliary tools in cotton breeding.
• To enhance the classification of cotton genotypes for improved fiber quality.
• To demonstrate the efficacy of ANNs in identifying superior cotton genotypes.

Main Methods:

• Utilized evaluation data from 40 cotton genotypes across two harvest seasons (2013/14, 2014/15).
• Trained ANNs using replicate data of 20 genotypes, focusing on key fiber quality traits (length, uniformity, strength, micronaire, etc.).
• Developed a fiber quality index based on weighted scores of HVI-evaluated characteristics.

Main Results:

• ANNs demonstrated a high capacity for correctly classifying genotypes based on the fiber quality index.
• Combining fiber length with short fiber index, fiber maturation, and micronaire index yielded superior classification results compared to using fiber length alone.
• Training ANNs with mean data from new genotypes, using models trained on replicate data, improved classification outcomes.

Conclusions:

• Artificial neural networks hold significant potential for application across various stages of cotton genetic improvement programs.
• ANNs can effectively aid in enhancing the fiber quality of future cotton cultivars.
• This computational intelligence approach offers a promising alternative/complement to traditional statistical methods in cotton breeding.