Phenotypic Structure in the Conus pennaceus Complex: Integrating Interpretable Morphometrics and Self-Supervised Image Embeddings

Published on: August 2026

0009-0002-9238-4007

This study investigates morphological variation within the Conus pennaceus complex using a dataset of 1,000 physical specimens. Conventional shell measurements were combined with self-supervised DINOv3 image embeddings representing shell shape, colour and surface pattern. The analyses examined morphological differentiation among accepted species, historical forms and C. pennaceus sensu stricto, while accounting for anatomical view, image provider and geography. Deep embeddings showed biologically interpretable correspondence with conventional morphometric traits, while also capturing broader multidimensional phenotype structure. Accepted species differed substantially in their degree of morphological separation, and the synonymized form C. elisae showed particularly strong differentiation. Shape and surface pattern were partly independent phenotype components, producing individual morphological mosaicism. Within C. pennaceus sensu stricto, variation was structured but did not provide evidence for recurrent discrete morphotypes. The results provide quantitative morphological hypotheses for subsequent integrative taxonomic validation.

The following links provide to the full technical document, a document with supplemental information, and shorter manuscript

Reports

Shorter manuscript
A condensed journal-style version presenting the principal methods, results and conclusions while referring to the supplementary information for technical and supporting detail.

Full Manuscript
The complete version of the publication, including the full methodological, analytical and interpretive detail.

Supplemental Methods and Results
Detailed supporting material, including extended methods, diagnostic analyses, complete statistical results, audit information and additional figures and tables.



Minimum number of images needed for each species

Minimum number of images needed for each species

Identifying Shells using Convolutional Neural Networks: Data Collection and Model Selection

Identifying Shells using Convolutional Neural Networks: Data Collection and Model Selection

Fine-tuning EfficientNetV2 Models

Fine-tuning EfficientNetV2 Models

Hierarchical CNN to identify Mollusca

Hierarchical CNN to identify Mollusca

Coping with the open-world problem when identifying Mollusca

Coping with the open-world problem when identifying Mollusca

AI in Integrative Taxonomy: Mollusk Species Identification with Deep Learning

AI in Integrative Taxonomy: Mollusk Species Identification with Deep Learning

A Large-Scale Convolutional Neural Network for Fine-Grained Conus Shell Identification

A Large-Scale Convolutional Neural Network for Fine-Grained Conus Shell Identification

Unveiling Morphological Insights in Biological Imagery Through CNN Interpretability Techniques

Unveiling Morphological Insights in Biological Imagery Through CNN Interpretability Techniques

Analyzing Intra- and Inter-Class Variability and Detecting Outliers in CNN Seashell Image Classification Models

Analyzing Intra- and Inter-Class Variability and Detecting Outliers in CNN Seashell Image Classification Models

Few-Shot Learning with a Pretrained CNN: Preliminary Results on Genus Morum (Harpidae)

Few-Shot Learning with a Pretrained CNN: Preliminary Results on Genus Morum (Harpidae)

Impact of Studio versus Field Imagery on CNN Performance for Mollusc Shell Analysis

Impact of Studio versus Field Imagery on CNN Performance for Mollusc Shell Analysis

Optimizing Molluscan CNN Taxonomy: Balancing Hierarchy Simplification, Data Volume, and Augmentation for Improved Classification

Optimizing Molluscan CNN Taxonomy: Balancing Hierarchy Simplification, Data Volume, and Augmentation for Improved Classification

A Deep Learning Workflow for Quantifying Shell Morphology: A Pilot Study on the Phylogenetics of Bufonaria

A Deep Learning Workflow for Quantifying Shell Morphology: A Pilot Study on the Phylogenetics of Bufonaria

Technical Reports for a Total-Evidence Phylogenetics of Terebridae

Technical Reports on Total-Evidence Phylogenetics of Terebridae

Building a New Family-Level Model: Why Generalization Matters

Building a New Family-Level Model: Why Generalization Matters

Species Delimitation

Species Delimitation as Human-Led Triage: A Stability-Aware Embedding Benchmark with COX1 Validation in Cone Snails ...

Epitonium CNN Models

Epitoniidae: Scientific Background, Hierarchical CNN Modelling, and Performance of One Family-Level and Six Genus-Level Models

Tellinidae CNN Models

Visual Diagnosability in Tellinidae (Mollusca: Bivalvia: Tellinoidea): Genus-First CNN Routing Improves Species-Level Shell Identification

YOLO-Assisted Object-Centric Preprocessing Improves Family-Level Shell Classification

YOLO-Assisted Object-Centric Preprocessing Improves Family-Level Shell Classification on Real-World User Uploads

Source-Aware Training Improves Robustness in Leakage-Aware CNN Species Classification of Morum Röding, 1798 (Gastropoda: Harpidae)

Source-Aware Training Improves Robustness in Leakage-Aware CNN Species Classification of Morum Röding, 1798 (Gastropoda: Harpidae)

Conservative Statistical Label Curation for Leakage-Aware CNN Species Classification of Harpa Röding, 1798

Conservative Statistical Label Curation for Leakage-Aware CNN Species Classification of Harpa Röding, 1798

DINOv3 Improves Class-Balanced Recovery but Not Global Accuracy in Harpa Shell Classification

DINOv3 Improves Class-Balanced Recovery but Not Global Accuracy in Harpa Shell Classification

Leakage-Aware and Audit-Governed CNN Classification of Clathurellidae (Conoidea; Gastropoda) Shell Images

Leakage-Aware and Audit-Governed CNN Classification of Clathurellidae (Conoidea; Gastropoda) Shell Images

Phenotypic Structure in the Conus pennaceus Complex: Integrating Interpretable Morphometrics and Self-Supervised Image Embeddings

Phenotypic Structure in the Conus pennaceus Complex: Integrating Interpretable Morphometrics and Self-Supervised Image Embeddings