chemotaxonomic survey of fatty acids in littoral algae from the persian gulf: applicationof machine learning for characterization of algae

The ocean covers 70 percent of earth s surface, and is the natural habitat of many plants, animals,and microorganisms. algae are some of the most common organisms inhabiting the earth [1]. thealgae group is divided into multicellular organisms, “macroalgae” or seaweed, and unicellularorganisms, known as “microalgae” (measuring from 1 µm to several cm). algae are an importantsource of vitamins, some essential minerals and trace elements, proteins, polyunsaturated fattyacids including omega-3 fatty acids, polysaccharides, polyphenols, sterols, pigments, amino acids,antioxidants, and fiber. algae have been used in many industries, including chemical, cosmetic,pharmaceutical, environmental cleaning, feed and fertilizer, conventional food, and fermentedfood. studies of algae biological activity demonstrated that they possess antioxidant, antibacterial,antiviral, and antifungal properties. among the various research fields in which macro- andmicroalgae are appearing, food technology is one of the most important areas. fatty acids (fa)are widely occurring in natural fats and dietary oils, and they are also critical nutritious substancesand metabolites in living organisms. degenerative diseases related to inappropriate fasconsumption cause two-thirds cases of the population death who are living in affluent,industrialized nations. fas and lipids are constituents of all algae cells. lipids represent 1–5% ofalgal dry matter and exhibit an interesting pufa composition. algal fas are beneficial and act asprophylactic supplements for type-2 diabetes, atherosclerosis, coronary heart diseases,arrhythmias, and cancer [2]. the aim of the present study is to conduct a chemotaxonomic surveyof fatty acids in littoral algae from the persian gulf. in this way, the fatty acids of littoral algaefrom the persian gulf (green, red, and brown algae) were derivatized into corresponding fatty acidmethyl esters (fames) and were analyzed by gas chromatography with flame ionization detector(gc-fid) instrument. machine learning methods, including linear discriminant analysis (lda),partial least squares discriminant analysis (pls-da), and support vector machine (svm), wereused to construct the models which extract significant variables, visualize discriminations, andclassify the studied algae samples based on their fatty acid fingerprints. the results demonstratedthat machine learning methods, including lda, pls-da, and svm, can characterize and classifythe macroalgae samples based on their fatty acid composition (the obtained accuracies for thecalibration and the test sets were between 98.0% and 100%).