Diamond, K. M. | Search bioRxiv

bioRxiv2021

Computational anatomy and geometric shape analysis enables analysis of complex craniofacial phenotypes in zebrafish mutants

Due to the complexity of fish skulls, previous attempts to classify craniofacial phenotypes have relied on qualitative features or 2D landmarks. In this work we aim to identify and quantify differences in 3D craniofacial phenotypes in adult zebrafish mutants. We first estimate a synthetic normative zebrafish template using microCT scans from a sample pool of wildtype animals using the Advanced Normalization Tools (ANTs). We apply a computational anatomy (CA) approach to quantify the phenotype of zebrafish with disruptions in bmp1a, a gene implicated in later skeletal development and whose human ortholog when disrupted is associated with Osteogenesis Imperfecta. Compared to controls, the bmp1a fish have larger otoliths and exhibit shape differences concentrated around the operculum, anterior frontal, and posterior parietal bones. Moreover, bmp1a fish differ in the degree of asymmetry. Our CA approach offers a potential pipeline for high throughput screening of complex fish craniofacial phenotypes, especially those of zebrafish which are an important model system for testing genome to phenome relationships in the study of development, evolution, and human diseases. Summary statementA computational anatomy approach offers a potential pipeline for high throughput screening of complex zebrafish craniofacial phenotypes, an important model system for the study of development, evolution, and human diseases.

developmental biology↗

bioRxiv2021

Hierarchy-guided Neural Networks for Species Classification

O_LISpecies classification is an important task that is the foundation of industrial, commercial, ecological, and scientific applications involving the study of species distributions, dynamics, and evolution. C_LIO_LIWhile conventional approaches for this task use off-the-shelf machine learning (ML) methods such as existing Convolutional Neural Network (ConvNet) architectures, there is an opportunity to inform the ConvNet architecture using our knowledge of biological hierarchies among taxonomic classes. C_LIO_LIIn this work, we propose a new approach for species classification termed Hierarchy-Guided Neural Network (HGNN), which infuses hierarchical taxonomic information into the neural networks training to guide the structure and relationships among the extracted features. We perform extensive experiments on an illustrative use-case of classifying fish species to demonstrate that HGNN outperforms conventional ConvNet models in terms of classification accuracy, especially under scarce training data conditions. C_LIO_LIWe also observe that HGNN shows better resilience to adversarial occlusions, when some of the most informative patch regions of the image are intentionally blocked and their effect on classification accuracy is studied. C_LI

evolutionary biology↗