Valentin, S. | Search bioRxiv

bioRxiv2025

Uncovering the genetic basis of agronomic traits in over 1,000 grapevine genotypes derived from a disease resistance breeding program

Breeding disease-resistant grapevines that retain agronomic performance under variable climates requires loci and predictions that transfer across related hybrid families. We analyzed 1,081 genotypes from 95 crosses within the French INRAE-ResDur breeding program for 13 phenology, yield and berry-composition traits evaluated at five sites from 2006 to 2024. We integrated within-family QTL mapping, kinship- and population-adjusted multiple-population QTL mapping in 772 progeny, and structure-aware GWAS in 899-968 individuals, depending on the trait, together with cross-environment, cross-trait and local genomic estimated breeding-value analyses. Genetic and phenotypic differentiation among families strongly affected locus detection. Of 76 family-QTL intervals, seven, representing six trait-region hypotheses, were locally concordant across all three mapping frameworks. The strongest recurrent evidence involved a chromosome-16 region for veraison and harvest date, where a localGEBV block at 14.69 Mb ranked first for veraison and second for harvest; chromosome-14 cluster traits and chromosome-1 compactness emerged as additional validation priorities. Cross-environment meta-analysis detected no common fixed-effect association at 5% FDR but revealed extensive heterogeneous evidence. Cross-trait analysis grouped 270 significant marker tests into 54 candidate multi-trait regions, without establishing biological pleiotropy. Population-adjusted localGEBV yielded a mean leave-one-population-out correlation of 0.470 between phenotypic BLUPs and genomic scores across traits. These results distinguish compact haplotype-validation targets from background- and environment-dependent signals, supporting a staged strategy that combines marker- assisted selection for validated recurrent regions with externally validated multi-trait, multi- environment genomic prediction for polygenic traits.

genetics↗

bioRxiv2019

Interpreting neural decoding models of working memory load using grouped model reliance

The application of machine learning algorithms for decoding psychological constructs based on neural data is becoming increasingly popular. However, there is a need for methods that allow to interpret trained decoding models, as a step towards bridging the gap between theory-driven cognitive neuroscience and data-driven decoding approaches. The present study demonstrates grouped model reliance as a model-agnostic permutation-based approach to this problem. Grouped model reliance indicates the extent to which a trained model relies on conceptually related groups of variables, such as frequency bands or regions of interest in electroencephalographic (EEG) data. As a case study to demonstrate the method, random forest and support vector machine models were trained on within-participant single-trial EEG data from a Sternberg working memory task. Participants were asked to memorize a sequence of digits (0-9), varying randomly in length between one, four and seven digits, where EEG recordings for working memory load estimation were taken from a 3-second retention interval. Present results confirm previous findings in so far, as both random forest and support vector machine models relied on alpha-band activity in most subjects. However, as revealed by further analyses, patterns in frequency and particularly topography varied considerably between individuals, pointing to more pronounced inter-individual differences than reported previously. Author summaryModern machine learning algorithms currently receive considerable attention for their predictive power in neural decoding applications. However, there is a need for methods that make such predictive models interpretable. In the present work, we address the problem of assessing which aspects of the input data a trained model relies upon to make predictions. We demonstrate the use of grouped model-reliance as a generally applicable method for interpreting neural decoding models. Illustrating the method on a case study, we employed an experimental design in which a comparably small number of participants (10) completed a large number of trials (972) over multiple electroencephalography (EEG) recording sessions from a Sternberg working memory task. Trained decoding models consistently relied on alpha frequency activity, which is in line with existing research on the relationship between neural oscillations and working memory. However, our analyses also indicate large inter-individual variability with respect to the relation between activity patterns and working memory load in frequency and topography. Taken together, we argue that grouped model reliance provides a useful tool to better understand the workings of (sometimes otherwise black-box) decoding models.

neuroscience↗