Abstract:Efficient subtype classification utilizing multiple biological information sources is clinically practical for precision treatment of inborn errors of metabolism (IEMs). This study utilized urinary GC-MS and dried blood spot MS/MS to enable complementary mass spectrometric characterization of pathological information for methylmalonic aciduria (MMA) subtypes. A novel mid-level data fusion strategy was subsequently described to effectively detect their subtle metabolic perturbations. In the proposed strategy, multivariate advantage was extracted via considering variable relationships in individual data blocks, separately. Further, it was extended to exploiting partial correlation across multiple data blocks, which induced the models between one data block and the first principal component (PC1) of another one for more valuable information recovery. The continuous values of PC1 replaced the traditional binary class labels, enabling suitable representation of disease heterogeneity numerically and subsequent accurate disease detection. Based on the unique role in exploratory data analysis, partial least squares discriminant analysis (PLS-DA) was coupled with bootstrap to form ensemble feature selection framework. It allowed the novel mid-level data fusion strategy to screen stable significant features. Investigated by two common MMA subtypes (mut and cblC) using urinary and blood metabolic profiles, the results showed that the novel mid-level data fusion strategy outperformed the traditional data fusion approaches both in predictive accuracy and biological interpretation. Besides, correlation network of the identified stable informative metabolites was visualized by chord plots, revealing different between-block biological interaction pattern for mut and cblC. Especially, in mut subtype, strong interactions could be occurred between the urinary metabolite of 3OH-propionic-2 and two blood metabolite ratios (C3/C2 and C3/C0), separately. The findings deepened the biological insights into the disease pathology of mut and cblC. The integration of multiple analytical sources combined with the proposed novel mid-level data fusion strategy thus opened the possibilities to achieve desirable subtype classification for MMA. It suitably guided the early and timely medical intervention for IEMs.

Enhancement of Species-Specific Analysis for Meat and Bone Meal by Matrix Fragments-Related Spectral Fusion

Rapid Determination of Componential Contents and Calorific Value of Selected Agricultural Biomass Feedstocks Using Spectroscopic Technology

[Prediction of Minced Pork Quality Attributes Using Visible and Near Infrared Reflectance Spectroscopy].

Improving Intramuscular Fat Assessment in Pork by Synergy Between Spectral and Spatial Features in Hyperspectral Image

Metabolomics Data Fusion Between Near Infrared Spectroscopy and High-Resolution Mass Spectrometry: A Synergetic Approach to Boost Performance or Induce Confusion.

Combination of an E-Nose and an E-Tongue for Adulteration Detection of Minced Mutton Mixed with Pork

[Association of Addison's disease and hyperthyroidism].

Qualitative and Quantitative Detection of Minced Pork Quality by Near Infrared Reflectance Spectroscopy

Development and evaluation of a meat mitochondrial metagenomic (3MG) method for composition determination of meat from fifteen mammalian and avian species

Clinicopathological predictors to predict sustained viral response rates in patients with chronic hepatitis C infection.

Identifying Meat from Grazing or Feedlot Yaks Using Visible and Near-infrared Spectroscopy with Chemometrics

Data fusion of near-infrared diffuse reflectance spectra and transmittance spectra for the accurate determination of rice flour constituents

Data Fusion by Ratio Modulation of Stable Isotope, Multi-Element, and Fatty Acids to Improve Geographical Traceability of Lamb

Discrimination of Boletaceae mushrooms based on data fusion of FT-IR and ICP–AES combined with SVM

Evaluation of Mutton Adulteration under the Effect of Mutton Flavour Essence Using Hyperspectral Imaging Combined with Machine Learning and Sparrow Search Algorithm

Fusion of three spectroscopic techniques for prediction of fatty acid in processed lamb

Determination of protein content of Auricularia auricula using near infrared spectroscopy combined with linear and nonlinear calibrations.

Near-Infrared Spectroscopy and Geostatistical Analysis for Modeling Spatial Distribution of Analytical Constituents in Bulk Animal By-Product Protein Meals

Classification of Fresh Meat Species Using Laser-induced Breakdown Spectroscopy with Support Vector Machine and Principal Component Analysis

Combined Relaxation Spectra for the Prediction of Meat Quality: A Case Study on Broiler Breast Fillets with the Wooden Breast Condition

Efficient Mass Spectrometric Characterization and Classification of Methylmalonic Aciduria Subtypes Through Urinary and Blood Metabolic Profiles Fusion