Genetic control of important yield attributing characters predicted through machine learning in segregating generations of interspecific crosses of tomato (Solanum lycopersicum L.)

Abstract

Skewness and kurtosis were analysed using mean data from the F₂ to F₅ generations of three interspecific tomato hybrids, incorporating two feral species: Solanum pimpinellifolium (Currant Tomato) and Solanum lycopersicum var. cerasiformae (Cherry Tomato). The study cantered on three crucial traits impacting fruit yield, with predictions generated through artificial neural networks and multiple linear regression. Plant height (PH), fruit weight (FW) and test weight of seeds (TSW) were identified as the most sensitive traits influencing fruit yield/plant in the Alisa Craig Aft × Solanum pimpinellifolium (Cross 1) and the Berika × Solanum lycopersicum var. cerasiformae (Cross 2). In contrast, fruits per plant (FPP), FW and TSW emerged as the key contributors to fruit yield in the BCT 115 dg × Solanum lycopersicum var. cerasiformae (Cross 3). Skewness and kurtosis distribution suggested complementary gene action with fewer number of segregating genes for PH in Cross 1, FW across all three cross combinations, TSW in Cross 1, and FPP in Cross 3. Duplicate gene action with fewer genes could be predicted for TSW in Cross 2 and Cross 3 while complementary gene action and a greater number of segregating genes were suggested for PH in Cross 2. Moderate-to-high narrow sense heritability was determined for all the characters suggesting phenotypic selection to be rewarding. Isolation of seven promising segregates based on the important yield attributers from three inter-specific hybrids in F₅ generation established the worth of advancing interspecific hybrids. This distinctive and novel breeding method offers exceptional potential for isolating superior tomato segregates through a targeted interspecific breeding programme.