Gaussian process regression for prediction and confidence analysis of fruit traits by near-infrared spectroscopy

Detection of fruit traits by using near-infrared(NIR)spectroscopy may encounter out-of-distribution samples that exceed the generalization ability of a constructed calibration model.Therefore,confidence analysis for a given prediction is required,but this cannot be done using common calibration models of NIR spectroscopy.To address this issue,this paper studied the Gaussian process regression(GPR)for fruit traits detection using NIR spectroscopy.The mean and variance of the GPR were used as the predicted value and confidence,respectively.To show this,a real NIR data set related to dry matter content measurements in mango was used.Compared to partial least squares regression(PLSR),GPR showed approximately 14%lower root mean squared error(RMSE)for the in-distribution test set.Compared with no confidence analysis,using the variance of GPR to remove abnormal samples made GPR and PLSR showed approximately 58%and 10%lower RMSE on the mixed distribution test set,respectively(when the type 1 error rate was set to 0.1).Compared with traditional one-class classification methods,the variance of the GPR can be used to effectively eliminate poorly predicted samples.

Ploidy and fruit trait variation in oil-tea Camellia:Implications for ploidy breeding

Plant polyploidy often occurs in conjunction with higher yield and superior quality.Therefore,obtaining polyploid germplasms is a significant part of breeding.The oil-tea Camellia tree is an important native woody plant that produces high-quality edible oil and includes many species of Camellia with different ploidies.However,whether higher ploidy levels in oil-tea Camellia trees are related to better traits remains unclear.In this study,the ploidy levels of 30 different oil-tea Camellia strains in three different species in the Sect.Paracamellia were determined by flow cytometry and chromosome preparation,and the phenotypic characteristics and fatty acid compositions of the fruits were examined by field observations and laboratory analyses.The correlations between the ploidy level of oil-tea Camellia and the main traits of the fruit were investigated.Our results showed that 10 Camellia lanceoleosa strains were diploid,10 Camellia meiocarpa strains were tetraploid and 10 Camellia oleifera strains were hexaploid.Hexaploid C.oleifera had larger fruit size and weight,more seeds per fruit,greater seed weight per fruit,higher oil content and greater yield per crown width than tetraploid C.meiocarpa and diploid C.lanceoleosa,but their fruit peel thickness and fresh seed rate were significantly lower,and these traits were significantly correlated with ploidy level.In addition,in terms of fatty acid composition,hexaploid C.oleifera had a higher oleic acid content than tetraploid C.meiocarpa and diploid C.lanceoleosa,but their linoleic acid,linolenic acid and arachidonic acid contents were lower.The contents of palmitic acid,stearic acid and total unsaturated fatty acids were not significantly correlated with ploidy level.In conclusion,certain correlations exist between the main characteristics of oil-tea Camellia fruit and the ploidy level,and increasing the ploidy level led to an increase in fruit yield with no effect on oil composition.The discovery of variations in the main characteristics of oil-tea Camellia fruit with different ploidies will facilitate germplasm innovation and lay a foundation for ploidy breeding and mechanistic research on fruit traits.

An assessment of the genetic diversity of pear(Pyrus L.)germplasm resources based on the fruit phenotypic traits

Germplasm resources are an important basis for genetic breeding and analysis of complex traits,and research on genetic diversity is conducive to the exploration and creation of new types of germplasm.In this study,the distribution frequency,coefficient of variation,Shannon-Wiener index,and variance and cluster analyses were used to analyze the diversity and trait differences of 39 fruit phenotypic traits from 570 pear accessions,which included 456 pear accessions from 11 species and 114 interspecific hybrid cultivars that had been stored in the National Germplasm Repository of Apple and Pear(Xingcheng,China).The comprehensive evaluation indices were screened by correlation,principal component and regression analyses.A total of 132 variant types were detected in 28 categorical traits of pear germplasm fruit,which indicate a rich diversity.The diversity indices in decreasing order were:fruit shape(1.949),attitude of calyx(1.908),flesh texture type(1.700),persistency of calyx(1.681),russet location(1.658),relief of area around eye basin(1.644),flavor(1.610)and ground color(1.592).The coefficient of variation of titratable acidity in the 11 numerical traits of pear germplasm fruit was as high as 128.43%,which could more effectively reflect the differences between pear accessions.The phenotypic differentiation coefficient V_(st)(66.4%)among the five cultivated pear species,including Pyrus bretschneideri(White Pear),P.pyrifolia(Sand Pear),P.ussuriensis(Ussurian Pear),P.sinkiangensis(Xinjiang Pear),and P.communis(European Pear),was higher than the within population phenotypic differentiation coefficient V_(st)(33.6%).The variation among populations was the main source of variation in pear fruit traits.A hierarchical cluster analysis divided the 389 accessions of six cultivated pear species,including P.pashia(Himalayan Pear),into six categories.There were certain characteristics within the populations,and the differences between populations were not completely clustered by region.For example,Sand Pear cultivars from Japan and the Korean Peninsula clustered together with those from China.Most of the White Pear cultivars clustered with the Sand Pear,and a few clustered with the Ussurian Pear cultivars.The Ussurian Pear and European Pear cultivars clustered separately.The Xinjiang Pear and Himalayan Pear did not cluster together,and neither did the cultivars.Seventeen traits,three describing fruit weight and edible rate(fruit diameter,fruit length and fruit core size),five describing outer quality and morphological characteristics(over color,amount of russeting,dot obviousness,fruit shape,and stalk length),and nine describing inner quality(flesh color,juiciness of flesh,aroma,flavor,flesh texture,flesh texture type,soluble solid contents,titratable acidity,and eating quality)were selected from the 39 traits by principal component and stepwise regression analyses.These 17 traits could reflect 99.3%of the total variation and can be used as a comprehensive evaluation index for pear germplasm resources.

Unravelling host-mediated effects on hemiparasitic Mexican mistletoe Psittacanthus calyculatus(DC.)G.Don traits linked to mutualisms with pollinators and seed dispersers

Aims While the influence of mistletoe-host interactions on each other’s evolution is well-recognized,the role of interactions between mis-tletoes and its mutualistic pollinators and seed dispersers mediated by host species is relatively unexplored.Methods Here,we examine the effects of host species(Mexican hawthorn Crataegus mexicana DC,black cherry Prunus serotina Ehrh.,leather-leaf Mexican oak Quercus crassipes Bonpl.)on flower morphology,nectar production,pollinator visitation rate and female reproduct-ive fitness in hemiparasitic Mexican mistletoe Psittacanthus caly-culatus(DC.)G.Don(Loranthaceae)populations at three different locations.We first measured the lengths of corolla,style,stamen,exerted stamen and anther,the length and width of the ovary and nectar availability for mistletoe flowers in natural populations.Then,we evaluated flower visitation and measured(length and width)and weighed the fruits and its seeds of mistletoes growing on each of the three host species.Finally,we evaluated the effects of host species,location and flower traits on fruit or seed size variation(both as proxies of reproductive fitness).Important Findings We found mistletoes growing on natural hosts P.serotina and Q.crassipes produced larger flowers,fruits and seeds than mistle-toes on cultivated C.mexicana.However,these differences varied across space.The amount of available nectar and hummingbird vis-itation rates in flowers of mistletoes on C.mexicana was higher at the three sites than in flowers of mistletoes on P.serotina or Q.cras-sipes.The effects of host species,study site and floral trait covariates affected significantly all fitness measures,indicating that mistletoes’reproductive fitness is affected differently depending on the host species and their site of occurrence.These host-associated differ-ences in reproduction might have implications for interactions with mutualistic vectors.