Horticulture, Environment, and Biotechnology Биотехнология

Predictive modeling of in vitro seed germination in Calanthe spp.: a comparative study of response surface methodology and machine learning

Lee, Han-Sol, Kim, Ye-Won, Murthy, Hosakatte Niranjana, Kim, Ja-Young, Yun, Kyung-Seop, Lee, Min-Jeong, Park, So-Young
Springer Nature · 2026-10-12 · DOI 10.1007/s13580-026-00853-9

Аннотация

Calanthe spp. belonging to the Orchidaceae family, possess high ornamental and medicinal value. However, their propagation efficiency under natural conditions remains low, and wild populations have been declining due to indiscriminate collection. To address this, we aimed to identify model-predicted culture conditions for in vitro seed germination rates of two Calanthe species. This study aimed to improve in vitro seed germination of Calanthe discolor and Calanthe striata f. sieboldii by evaluating the effects of 1-naphthaleneacetic acid (NAA), 6-benzyladenine (BA), and coconut water (CW), and by comparing response surface methodology (RSM) with machine-learning regression models. Seeds were cultured on Murashige and Skoog medium containing NAA (0–0.5 mg/L), BA (0–1.0 mg/L), and CW (0–10%, v/v) according to an augmented simplex-centroid design. The highest observed germination of C. discolor was obtained with 0.5 mg/L NAA, while RSM predicted a maximum germination of 1.68% at 0.758 mg/L BA and 2.42% CW. For C. striata f. sieboldii , the highest observed germination was obtained with 10% CW, which was also identified by RSM as the predicted optimum, with a germination response of 4.57%. The RSM models adequately described the species-specific response patterns within the investigated ranges, whereas the machine-learning models showed variable performance because of the limited dataset. These findings provide useful candidate medium conditions for improving asymbiotic seed germination of the two Calanthe species and establish a basis for further refinement and validation.

Читать у издателя →