Thamnaconus septentrionalis is an economically important marine aquaculture species in China. However, the acceptance rate of formulated feeds in commercial farming is only 30–40%, substantially lower than the 80–90% achieved with fresh feeds, which severely constrains the intensive development of this industry. The gut microbiota-mediated regulatory mechanisms underlying the effects of different feed types on growth performance remain unclear, limiting the precise development of efficient formulated feeds. This study established four feed types (commercial pellet feed K, custom-formulated feed P, frozen shrimp X, and fresh fish meat Y) through a 60-day feeding trial. Growth performance data, 16S rRNA sequencing, and untargeted metabolomics were analyzed. Random Forest-Partial Least Squares Regression models were employed to identify key microbial-metabolite features. Results indicated that the Y group exhibited the optimal feed conversion ratio (1.14), with intestinal Firmicutes abundance (45.3%) significantly higher than the K group (28.5%). Short-chain fatty acid levels increased by more than 350-fold, enriching short-chain fatty acid-producing bacteria such as Lactobacillus and Faecalibacterium. The P group, formulated with high fishmeal content (40%), achieved performance levels comparable to the Y group across most indicators. Machine learning models identified key microbial-metabolite features predicting growth performance, providing a multi-omics framework for developing efficient formulated feeds for marine carnivorous fish.

Thamnaconus modestus (black scraper) is an economically important fish species in Chinese coastal fisheries, yet its pronounced fishy off-odor, primarily attributed to sulfur-containing compounds and trimethylamine (TMA), severely limits consumer acceptance and product diversification. However, a systematic investigation into how different thermal processing methods affect its volatile flavor profile is lacking. This study employed an integrated multi-instrumental flavoromics platform combining sensory evaluation, electronic nose (E-nose), electronic tongue (E-tongue), gas chromatography–ion mobility spectrometry (GC-IMS), and headspace solid-phase microextraction gas chromatography–mass spectrometry (HS-SPME-GC-MS), coupled with chemometric analysis, to systematically characterize the aroma variations of T. modestus subjected to steaming, boiling, deep-frying, and roasting treatments compared with raw samples. A total of 62 (GC-IMS) and 129 (GC-MS) volatile compounds were identified, from which 78 characteristic markers (VIP > 1) and 45 key odorants (OAV ≥ 1) were screened. Thermal processing markedly reduced sulfur-containing compounds and TMA concentrations (raw >> steamed ≈ boiled >> deep-fried > roasted) while promoting lipid oxidation- and Maillard reaction-derived aldehydes and furans. Two distinct flavor modulation patterns were revealed: moist-heat methods (steaming, boiling) generated grassy/fatty notes through moderate lipid oxidation, whereas dry-heat methods (deep-frying, roasting) produced characteristic roasted/nutty notes via synergistic activation of Strecker degradation and Maillard reaction. These findings provide scientific evidence for precise flavor quality control and diversified processing optimization of T. modestus products.