Regional feature extraction of various fishes based on chemical and microbial variable selection using machine learning

Abstract

We introduce a method for extracting regional and habitat features of various fish species based on chemical and microbial correlations that incorporate integrated analysis and a variable selection approach. We characterized 24 fish species from two marine regions in Japan, in terms of the metabolic and inorganic profiles of muscle and gut contents, as well as gut microbes. Using machine learning, the integrated analysis based on the metabolic, inorganic, and microbial profiles of muscle and gut contents allows the characterization of both the fish species and habitat regions. The results revealed that the fish muscle tissue profile provides high-value data for evaluating ecosystems and discriminating fish populations based on species and regions. To visualize the regionality and habitat, we developed a method to efficiently extract the most important variables using the machine learning approach, followed by correlation analysis of variations in muscle and gut content profiles. The correlation networks enabled efficient visualization of marine ecosystems in the Tohoku and Kanto regions of Japan. This method should be useful for evaluating fish habitats and elucidating associated environmental chemical networks.

Fish ecotyping based on machine learning and inferred network analysis of chemical and physical properties

“…3 Graduate School of Bioagricultural Sciences and School of Agricultural Sciences, Nagoya University, 1 Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan. * email: jun.kikuchi@riken.jp chemical separation, provides an ideal approach to explore the complicated interactions of genes, growth stages and the surrounding environment of wild fish in their natural state [10][11][12][13] . Advances in data science and big data technologies, such as machine learning, deep learning and artificial intelligence, have enabled the extraction of useful information by evaluating the relative importance of each feature according to model coefficients and the establishment of simulation models to predict ecosystem dynamics on the basis of high dimensionality and large data sets in which raw data are largely unlabeled and uncategorized 12,[14][15][16] .…”

Signal Deconvolution and Noise Factor Analysis Based on a Combination of Time–Frequency Analysis and Probabilistic Sparse Matrix Factorization

“…Briefly, 1 H-NMR data were by recorded using an Avance II 700 Bruker spectrometer equipped with a 5-mm inverse CryoProbe operating at 700.153 MHz for 1 H. In the 1 H -NMR data, the number of data using CPMG pulse sequence was 2386, the number of data using WATERGATE pulse sequence was 2760, and the number of data in the 1D LED experiment using bipolar gradients (diffusion-edited) pulse sequence was 975 [58][59][60][61]. Regarding these large data sets, a summary of information on the sample and acquisition parameters (the sample title, solvent, acquisition time, acquisition point, and the original SNR) is available at http://dmar.riken.jp/NMRinformatics/.…”

Large-Scale Evaluation of Major Soluble Macromolecular Components of Fish Muscle from a Conventional 1H-NMR Spectral Database

“…A total of 800 fish samples were sampled from June 2011 to February 2017 at multiple sites around Japan (Table S1) [4,[24][25][26][27][28]. After dissection, the fish muscle samples were freeze-dried and crushed into powder.…”