Marine biomass composition analysis traditionally requires time-consuming processes and domain expertise. This study demonstrates the effectiveness of rapid evaporative ionization mass spectrometry (REIMS) combined with advanced machine learning (ML) techniques for accurate marine biomass composition determination. Using fish species and body parts as model systems representing diverse biochemical profiles, we investigate various ML methods, including unsupervised pretraining strategies for transformers. The deep learning approaches consistently outperformed traditional machine learning across all tasks. For fish species classification, the pretrained transformer achieved 99.62% accuracy, and for fish body parts classification, the transformer achieved 84.06% accuracy. We further explored the explainability of the best-performing and predominantly black box models using local interpretable model-agnostic explanations and gradient-weighted class activation mapping to identify the important features driving the decisions behind each of the best performing classifiers. REIMS analysis with ML can be an accurate and potentially explainable technique for automated marine biomass composition analysis. Thus, REIMS analysis with ML has potential applications in quality control, product optimization, and food safety monitoring in marine-based industries.

The batch detection of marine biomass constitutes a significant real-world application within the fish processing industry, contributing to food safety, fraud prevention, and stock management. Recent advancements have demonstrated that Rapid Evaporative Ionization Mass Spectrometry (REIMS) when coupled with Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA), yields exceptional outcomes in fraud detection, contamination identification, and biomass analysis. Although several studies have employed REIMS and OPLS-DA for species identification and contamination detection—including limited applications to marine biomass—these efforts have not yet addressed the challenge of batch detection, which involves determining the specific batch of processed samples from which a fish originates. Contrastive Learning, an emerging alternative to conventional binary classification, has proven effective for batch detection of marine biomass analyzed via REIMS. Leveraging a high-dimensional REIMS dataset provided by Plant and Food Research, New Zealand, comprising mass spectrometry profiles of New Zealand marine biomass, we propose a novel Contrastive Learning approach termed SpectroSim, building upon the SimCLR framework. The new method introduces a bespoke encoder head, replacing the traditional ResNet backbone with a Transformer architecture, alongside a custom projection head meticulously designed for mass spectrometry data. Comprehensive experimental results indicate that SpectroSim surpasses the balanced classification accuracy of established deep learning frameworks and other prevalent baseline models. Notably, as an unsupervised methodology, SpectroSim achieves near-perfect accuracy (98.02%) in a self-supervised context, independent of class labels.

Fish is approximately 40% edible fillet. The remaining 60% can be processed into low-value fertilizer or high-value pharmaceutical-grade omega-3 concentrates. High-value manufacturing options depend on the composition of the biomass, which varies with fish species, fish tissue and seasonally throughout the year. Fatty acid composition, measured by Gas Chromatography, is an important measure of marine biomass quality. This technique is accurate and precise, but processing and interpreting the results is time-consuming and requires domain-specific expertise. The paper investigates different classification and feature selection algorithms for their ability to automate the processing of Gas Chromatography data. Experiments found that SVM could classify compositionally diverse marine biomass based on raw chromatographic fatty acid data. The SVM model is interpretable through visualization which can highlight important features for classification. Experiments demonstrated that applying feature selection significantly reduced dimensionality and improved classification performance on high-dimensional low sample-size datasets. According to the reduction rate, feature selection could accelerate the classification system up to four times.

Navigating the analysis of mass spectrometry data for marine biomass and fish demands a technologically adept approach to derive accurate and actionable insights. This research will introduce a novel AI methodology to interpret a substantial repository of mass spectrometry datasets, utilizing pre-training strategies like Next Spectra Prediction and Masked Spectra Modeling, targeting enhanced interpretability and correlation of spectral patterns with chemical attributes. Three core research objectives are explored: 1) precise fish species and body part identification via binary and multi-class classification, respectively; 2) quantitative contaminant analysis employing multi-label classification and multi-output regression; and 3) traceability through pair-wise comparison and instance recognition. By validating against traditional baselines and various downstream tasks, this work aims to enhance chemical analytical processes and offer fresh insights into the chemical and traceability aspects of marine biology and fisheries through advanced AI applications.