PENERAPAN PEMBELAJARAN MENDALAM UNTUK ANALISIS EKSPRESI GEN KETAHANAN KEKERINGAN PADA TANAMAN PADI LOKAL KAB. NIAS (Oryza sativa L.)

Abstrak

Rice plant (Oryza sativa L.) is a strategic food commodity that faces serious challenges due to the increasing occurrence of drought, especially in the Nias Islands region which has uneven distribution of rice field land and long dry seasons. This study aims to analyze the morpho-physiological response of local Nias rice to drought stress and develop a deep learning model to predict drought resistance gene expression indirectly through image data and physiological parameters. The study was conducted at the Faculty of Science and Technology, Universitas Nias, involving five local rice varieties. Physiological analysis included measurement of proline content, malondialdehyde (MDA), and relative water content (RWC), while plant images were processed using a Convolutional Neural Network (CNN) architecture based on ResNet50. The results showed that varieties originating from Bawolato and Idanogawo had higher resistance with an increase in proline up to 2.8 times and RWC stability above 70%. The developed deep learning model was able to predict the level of drought resistance gene expression with an accuracy of 91.3%, and identify leaf visual characteristics as the main indicators of water stress. This approach proves the great potential of integrating phenotypic data and artificial intelligence in accelerating the identification of drought-resistant varieties efficiently and non- destructively, supporting the development of precision agriculture in the Nias region.