NIRSpredict: a platform for predicting plant traits from near infra-red spectroscopy

Area of Science:

• Plant Science
• Ecology
• Bioinformatics

Background:

• Near-infrared spectroscopy (NIRS) is increasingly used for plant phenotyping.
• Predicting diverse plant traits from spectral data remains challenging.

Purpose of the Study:

• To develop a user-friendly application, Shiny NIRSpredict, for predicting numerous plant traits using NIRS data.
• To provide access to a curated database of Arabidopsis thaliana phenotypic traits.

Main Methods:

• Developed deep learning models for trait prediction from NIRS data.
• Integrated models into a Shiny web application for accessibility.
• Compiled and curated a database of 81 Arabidopsis thaliana phenotypic traits.

Main Results:

• NIRSpredict accurately predicts 81 phenotypic traits, including functional traits and chemical compounds, from NIRS values.
• The application offers functionalities for trait prediction, database exploration, and user data submission.
• Provides a valuable resource for phenomic analysis in plant science.

Conclusions:

• Shiny NIRSpredict offers an efficient and accessible platform for NIRS-based plant trait prediction.
• The tool facilitates phenomic research in functional and evolutionary ecology by characterizing plant populations.
• Enhances the adoption of phenomics through easy access to trait data and prediction models.