Cecal microbiota and metabolomic profiles associated with residual feed intake divergence in laying ducks Authors: N. Li, J. Yu, J-G. Wang, H. Shen, J. Wang Published: 2026 ABSTRACT Feed efficiency is an important determinant of productivity in laying ducks, but the cecal microbiota and metabolomic profiles associated with residual feed intake (RFI) divergence remain unclear. In this study, 767 laying ducks were phenotyped for RFI, and cecal samples from low-RFI (LRFI) and high-RFI (HRFI) ducks (n = 20 per group) were subjected to 16S rRNA gene sequencing and LC–MS-based untargeted metabolomics. LRFI ducks consumed 25.5% less feed and showed lower RFI and feed conversion ratio than HRFI ducks; however, they also exhibited higher average daily egg mass and egg production rate, whereas body weight-related traits did not differ between groups. Cecal microbiota composition and overall metabolomic profiles differed clearly between LRFI and HRFI ducks. At the genus level, Phocaeicola and Olsenella were enriched in LRFI ducks, and 251 differential metabolites with varying annotation-confidence levels were detected between groups. Pathway analysis descriptively indicated recurring themes related to lipid metabolism and bile acid/steroid-associated metabolism, although no KEGG pathway remained significant after multiple-testing correction. Integrated analysis further showed exploratory correlation patterns among differential genera, metabolites, feed-efficiency-related traits, and laying-performance traits. These findings identify cecal microbial and metabolic features associated with the overall LRFI–HRFI phenotypic contrast and provide a basis for targeted validation in larger laying-duck populations. Introduction Feed represents one of the largest cost components in modern poultry production, accounting for approximately 70% of variable costs in many systems (Zampiga et al., 2018; Zeng et al., 2018; Gao et al., 2025). Improving feed efficiency is therefore important not only for reducing production costs, but also for improving resource use and lowering environmental outputs such as nitrogen excretion (Sell-Kubiak et al., 2017; Zeng et al., 2018; Sun et al., 2023; Bernard et al., 2024). In laying ducks (Anas spp.), nutrients must simultaneously support somatic maintenance and egg formation during the laying period (van Eck et al., 2023). Notably, substantial inter-individual variation in feed intake persists among birds with comparable body weight and egg output (Sun et al., 2023; Wan et al., 2025), suggesting that feed efficiency reflects not only production level but also intrinsic differences in nutrient utilization and metabolic regulation. Residual feed intake (RFI), defined as the deviation between observed and predicted feed intake required for maintenance and productive output, has emerged as a robust metric for evaluating feed efficiency independent of production traits (Wen et al., 2018). Compared with ratio-based indices such as feed conversion ratio (FCR), RFI better captures inherent differences in metabolic efficiency and nutrient utilization, with reduced confounding from body size or production level (Bai et al., 2022; Liu et al., 2024; Ramankevich et al., 2025). In laying ducks, RFI is moderately heritable and largely independent of egg production traits such as egg mass (Bai et al., 2022), and divergence in RFI does not necessarily compromise egg quality (Zeng et al., 2018). Taken together, these characteristics make RFI a useful trait for studying biological differences related to feed-efficiency variation, especially those involving nutrient utilization and energy partitioning (Marett et al., 2017; Silva et al., 2020). Growing evidence suggests that gut microbiota is associated with feed efficiency variation in poultry (Metzler-Zebeli et al., 2019; Wen et al., 2021; Xu et al., 2021; Zhang et al., 2022; Sun et al., 2023; Bernard et al., 2024). The cecum is a major site of microbial fermentation and is important for nutrient utilization and host physiology (Wen et al., 2021; Du et al., 2025). Microbiota-derived metabolites, especially short-chain fatty acids (SCFAs), have been proposed as one possible link between microbial composition and host feed efficiency (Den Besten et al., 2013; He et al., 2023). In addition to SCFAs, gut microbiota is also involved in bile acid and lipid metabolism (Ramírez-Pérez et al., 2017; Yao et al., 2022; Collins et al., 2023), which are both related to host energy utilization. Together, these studies support a potential association between gut microbiota, the intestinal metabolic environment, and feed-efficiency variation. In poultry, microbial communities have been reported to explain part of the variation in residual feed intake (RFI) (Wen et al., 2021; Sun et al., 2023; Chang et al., 2024), and low-RFI ducks may harbor microbiota with greater metabolic potential, including a higher potential for SCFA production (Sun et al., 2023). Untargeted metabolomics can help describe metabolic differences associated with microbiota variation and feed efficiency (Lamichhane et al., 2018; Wu et al., 2019; Yao et al., 2022; Xie et al., 2025). However, several questions remain in laying ducks. Integrative studies that jointly characterize cecal microbiota, metabolomic profiles, and RFI divergence are still limited. In addition, most available evidence is based on association analyses, making it difficult to determine the directionality and biological significance of the relationships between microbiota-related metabolic features and feed-efficiency variation. Differences in sampling site, laying stage, and population background across studies further complicate the identification of consistent and biologically interpretable microbial and metabolic signatures (Wen et al., 2021; Zhou et al., 2023; Bernard et al., 2024). Therefore, the relationships between gut microbiota-associated metabolic features and RFI divergence in laying ducks remain to be clarified. Based on these considerations, we hypothesized that RFI divergence in laying ducks would be associated with differences in the cecal microbiota and metabolic environment. To investigate this, we combined RFI-based phenotyping with cecal 16S rRNA gene sequencing and LC–MS-based untargeted metabolomics to characterize microbial and metabolic features associated with feed-efficiency divergence in a uniform population of laying ducks. Therefore, this study aimed to compare production traits, cecal microbiota, and cecal metabolomic profiles between low- and high-RFI ducks, and to identify candidate microbial and metabolic features associated with feed-efficiency variation. Materials and methods Animal ethics All procedures were approved by the Institutional Animal Care and Use Committee of Shihezi University (Authorization No. A2025-545) and were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and the ARRIVE guidelines. Euthanasia procedures followed the AVMA Guidelines for the Euthanasia of Animals. As the primary institution responsible for this study, Jiangsu Agri-animal Husbandry Vocational College supervised the research and ensures adherence to these ethical standards. RFI divergence selection and cecal sample collection Animal Housing and Experimental Design. A total of 815 female Jinding ducks at 203 d of age were obtained from the National Waterfowl Germplasm Resource Pool. Ducks were housed individually in three-tier cages (45 × 32 × 43/38 cm; length × width × front/back height). Each cage was equipped with a feeder (28 × 10 × 10 cm) and a nipple drinker (one nipple per duck). Ducks had ad libitum access to feed and water throughout the experiment. All ducks were fed the same commercial basal diet purchased from Cofco Feed Co., Ltd. (Chaohu, China). The ingredient composition and calculated nutrient levels of the basal diet are presented in Table 1. Crude protein, calcium, total phosphorus, methionine, and lysine contents were determined according to AOAC procedures, whereas other nutrient values were calculated based on feed composition tables published in China Feed.Table 1Ingredients and nutrient levels of the diet (air-dry basis).IngredientsComposition, %NutrientLevels2Corn grain41.275Metabolizable energy, MJ/kg11.69Soybean meal (CP 43%)20.8Crude protein, %18.07Polishing powder20.0Ether extract, %4.95Corn gluten meal (CP 60%)4.00Crude fiber, %2.74Limestone powder8.50Calcium, %3.56Calcium hydrophosphate0.80Total phosphorus, %0.54Soybean oil2.582Methionine, %0.48DL-Methionine0.19Lysine, %1.12L-Lysine0.485Met + Cys, %0.77Salt0.29Threonine, %0.70L-Threonine0.078Tryptophan, %0.20Premix11.00Total100.001The premix provided per kilogram diet: vitamin A, 8000 IU; vitamin D3, 2500 IU; vitamin E, 50 IU; vitamin K3, 2.5 mg; vitamin B1, 4 mg; vitamin B2, 6 mg; vitamin B6, 6 mg; vitamin B12, 0.2 mg; nicotinic acid, 15 mg; D-pantothenic acid, 20 mg; folic acid, 1 mg; biotin, 0.2 mg; Fe, 75 mg; Cu, 10 mg; Mn, 125 mg; Zn, 115 mg; I, 0.3 mg; Se, 0.25 mg; choline, 800 mg; phytase, 1,000 U.2Crude protein, calcium, total phosphorus, methionine, and lysine contents of the basal diet were analyzed according to AOAC procedures. Other nutrient levels were calculated based on the feed ingredient composition and the nutrient composition values reported in the feed composition tables published in China Feed. The same batch of basal diet was used throughout the experimental period. Table 1Ingredients and nutrient levels of the diet (air-dry basis).IngredientsComposition, %NutrientLevels2Corn grain41.275Metabolizable energy, MJ/kg11.69Soybean meal (CP 43%)20.8Crude protein, %18.07Polishing powder20.0Ether extract, %4.95Corn gluten meal (CP 60%)4.00Crude fiber, %2.74Limestone powder8.50Calcium, %3.56Calcium hydrophosphate0.80Total phosphorus, %0.54Soybean oil2.582Methionine, %0.48DL-Methionine0.19Lysine, %1.12L-Lysine0.485Met + Cys, %0.77Salt0.29Threonine, %0.70L-Threonine0.078Tryptophan, %0.20Premix11.00Total100.001The premix provided per kilogram diet: vitamin A, 8000 IU; vitamin D3, 2500 IU; vitamin E, 50 IU; vitamin K3, 2.5 mg; vitamin B1, 4 mg; vitamin B2, 6 mg; vitamin B6, 6 mg; vitamin B12, 0.2 mg; nicotinic acid, 15 mg; D-pantothenic acid, 20 mg; folic acid, 1 mg; biotin, 0.2 mg; Fe, 75 mg; Cu, 10 mg; Mn, 125 mg; Zn, 115 mg; I, 0.3 mg; Se, 0.25 mg; choline, 800 mg; phytase, 1,000 U.2Crude protein, calcium, total phosphorus, methionine, and lysine contents of the basal diet were analyzed according to AOAC procedures. Other nutrient levels were calculated based on the feed ingredient composition and the nutrient composition values reported in the feed composition tables published in China Feed. The same batch of basal diet was used throughout the experimental period. Ingredients and nutrient levels of the diet (air-dry basis). The premix provided per kilogram diet: vitamin A, 8000 IU; vitamin D3, 2500 IU; vitamin E, 50 IU; vitamin K3, 2.5 mg; vitamin B1, 4 mg; vitamin B2, 6 mg; vitamin B6, 6 mg; vitamin B12, 0.2 mg; nicotinic acid, 15 mg; D-pantothenic acid, 20 mg; folic acid, 1 mg; biotin, 0.2 mg; Fe, 75 mg; Cu, 10 mg; Mn, 125 mg; Zn, 115 mg; I, 0.3 mg; Se, 0.25 mg; choline, 800 mg; phytase, 1,000 U. Crude protein, calcium, total phosphorus, methionine, and lysine contents of the basal diet were analyzed according to AOAC procedures. Other nutrient levels were calculated based on the feed ingredient composition and the nutrient composition values reported in the feed composition tables published in China Feed. The same batch of basal diet was used throughout the experimental period. Room temperature was maintained between 18 and 26°C under a 16 h light/8 h dark photoperiod throughout the experimental period. During the preliminary screening period from 203 to 210 d of age, egg production was recorded daily and body weight (BW) was measured at 210 d. Ducks with leg disorders (n = 2) or non-laying status, defined as an egg production rate of 0% during the screening period (n = 41), were excluded. The remaining 772 ducks were enrolled in a 56-d residual feed intake (RFI) test conducted from 210 to 266 d of age. Five ducks that died during the experimental period were excluded from subsequent analyses, leaving 767 ducks for final RFI calculation. Throughout the RFI test, feed offered, residual feed, spilled feed, egg production, egg weight, initial BW, and final BW were recorded on an individual basis. To calculate metabolic body weight (MBW), body weight gain (BWG), average daily gain (ADG), average daily feed intake (ADFI), average daily egg mass (ADEM), egg production rate (EPR), feed conversion ratio (FCR), residual feed intake (RFI). Residual Feed Intake Calculation and Divergent Group Selection. Residual feed intake (RFI) was used to evaluate feed efficiency and was calculated using the following multiple linear regression model:RFI=ADFI−(u+a×MBW+b×ADG+c×ADEM),where ADFI represents average daily feed intake, u is the intercept, a, b, and c are partial regression coefficients, MBW is metabolic body weight, ADG is average daily gain, and ADEM is average daily egg mass. Metabolic body weight was calculated as:MBW=[(initialBW+finalBW)/2]0.75. Ducks were ranked according to RFI values. The 50 ducks with the lowest RFI values were identified as the low residual feed intake (LRFI) divergent pool, and the 50 ducks with the highest RFI values as the high residual feed intake (HRFI) divergent pool. All ducks in the LRFI divergent pool had RFI values below mean − 0.5 standard deviation (SD), whereas all ducks in the HRFI divergent pool had RFI values above mean + 0.5 SD. From each divergent pool, the 20 ducks with the most extreme RFI values—the 20 lowest-RFI ducks in the LRFI pool and the 20 highest-RFI ducks in the HRFI pool—were selected for paired 16S rRNA gene sequencing and untargeted metabolomics to maximize the phenotypic contrast between groups. Cecal Content Collection. Cecal contents were collected from the 40 ducks selected using the extreme-RFI criteria described above. At the end of the experiment, the selected ducks were euthanized by CO₂ asphyxiation followed by cervical dislocation according to the AVMA Guidelines for the Euthanasia of Animals. Approximately 0.5 to 1.0 g of cecal contents was aseptically collected from each bird, immediately frozen in liquid nitrogen, and stored at −80°C until microbial DNA extraction and untargeted metabolomics analysis. 16S rRNA gene sequencing and microbiota analysis Genomic DNA was extracted from cecal contents using the MagBeads Fast DNA Kit (MP Biomedicals, CA, USA) according to the manufacturer’s instructions. DNA concentration and quality were assessed using a NanoDrop NC2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and 1% agarose gel electrophoresis. The V3–V4 region of the bacterial 16S rRNA gene was amplified using primers 338F (5′-ACTCCTACGGGAGGCAGCA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Purified amplicons were pooled in equimolar concentrations and sequenced on an Illumina NovaSeq platform (2 × 250 bp paired-end). Raw paired-end reads were processed using QIIME2 (v2019.4). Quality filtering, denoising, paired-end merging, and chimera removal were performed using the DADA2 pipeline to generate amplicon sequence variants (ASVs). Singleton ASVs were removed before downstream analyses. Per-sample sequence counts retained after each processing step and observed ASV counts are provided in Supplementary Table S1. Before alpha- and beta-diversity analyses, the singleton-filtered ASV table was rarefied to 22,422 sequences per sample, corresponding to approximately 95% of the minimum library size after singleton removal, and all 40 samples were retained. Taxonomic classification was performed against the Greengenes2 database (release 2022.10) using a Naive Bayes classifier trained on the V3–V4 region. Alpha diversity indices were calculated at the ASV level. Beta diversity was evaluated by principal coordinate analysis (PCoA) based on Bray–Curtis dissimilarities. Differential bacterial taxa between groups were explored using linear discriminant analysis effect size (LEfSe), with a linear discriminant analysis (LDA) score threshold of 2.5. Untargeted metabolomics analysis Cecal content samples were processed for untargeted metabolomics analysis following a standardized protocol. Briefly, approximately 50 mg of sample was extracted with methanol containing 2-amino-3-(2-chlorophenyl)propionic acid as an internal standard. After homogenization, ultrasonic extraction, and centrifugation, supernatants were filtered through a 0.22 μm membrane and transferred to LC–MS vials. Metabolite separation was performed using a Vanquish UHPLC system (Thermo Fisher Scientific, USA) equipped with an ACQUITY UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm; Waters, Milford, MA, USA). Mass spectrometry was conducted on an Orbitrap Exploris 240 mass spectrometer (Thermo Fisher Scientific, USA) operated in both positive and negative ion modes. Raw LC–MS data were converted into mzXML format using MSConvert and processed using the XCMS package in R for peak detection, retention-time correction, and feature alignment. Following missing-value filtering, quality-control (QC) variability filtering, and other QC procedures, 1,828 ion features in the positive-ionization mode and 1,094 ion features in the negative-ionization mode were retained in the processed quantitative matrices for subsequent statistical analysis before metabolite annotation. Metabolite annotation was performed by matching accurate mass, retention time, and MS/MS fragmentation spectra against public databases and spectral libraries, including HMDB, MassBank, KEGG, LIPID MAPS, and mzCloud, together with the analytical provider’s in-house reference library. Annotation confidence was classified into three levels according to the evidence-based criteria provided by the analytical service provider. Level 1 compounds were confidently identified by matching retention time, accurate mass, and MS/MS fragmentation patterns against an in-house reference library established using authentic chemical standards. Level 2 compounds were putatively annotated by accurate-mass and MS/MS spectral matching against public databases or spectral libraries without authentic-standard-library confirmation. Level 3 compounds were tentatively annotated based on lower-confidence spectral or database evidence. Principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and orthogonal partial least squares discriminant analysis (OPLS-DA) were performed using the ropls package in R. Differential metabolites were identified based on variable importance in projection (VIP) > 1, fold change (FC) > 1.5 or < 0.67, and P < 0.05. Pathway enrichment analysis was conducted using MetaboAnalyst based on the KEGG database. Only features identified as differential between the LRFI and HRFI groups were included in the integrated correlation analyses. Specifically, the 15 genera retained after Wilcoxon rank-sum testing with Benjamini–Hochberg correction (FDR-adjusted P < 0.05) and metabolites meeting VIP > 1, FC > 1.5 or < 0.67, and nominal P < 0.05 were included. Spearman rank correlations were calculated for differential genus–trait, differential metabolite–trait, and differential genus–metabolite pairs. Associations meeting |ρ| ≥ 0.5 and nominal P < 0.05 were displayed to focus the visualization on stronger pairwise relationships and reduce network complexity. Statistical analysis Residual feed intake was calculated based on the 767 ducks that completed the RFI test. After divergent RFI selection, the 20 ducks with the most extreme RFI values from each divergent pool were used for all subsequent analyses in the present study, including production traits, feed-efficiency traits, 16S rRNA gene sequencing, and untargeted metabolomics. This study was an observational comparison of naturally divergent RFI phenotypes within a single managed population of laying ducks. Ducks were housed individually throughout the experimental period; therefore, the individual duck was considered both the experimental unit and the observational unit for production traits and omics analyses. No formal a priori power calculation was performed for the omics analyses because suitable effect-size estimates from comparable laying-duck populations were unavailable. As a general univariate benchmark, 20 ducks per group provided 80% power to detect a standardized between-group difference of approximately 0.91 at α = 0.05; however, the power of 16S rRNA gene sequencing and metabolomics analyses also depends on feature variability, inter-individual dispersion, and multiple-testing correction. Residual feed intake was estimated using the REG procedure of SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) based on a multiple linear regression model, with average daily feed intake as the dependent variable and metabolic body weight, average daily gain, and average daily egg mass as explanatory variables. The residual from the regression model was defined as individual RFI. Model assumptions were assessed by examining residual-versus-fitted plots and normal Q–Q plots. Potentially influential observations were evaluated using studentized residuals and Cook’s distance. No marked violations of linearity, homoscedasticity, or residual normality were observed, and no individual observation exerted a substantial influence on the fitted model. Therefore, all 767 ducks that completed the RFI test were retained for RFI estimation. Data normality was assessed using PROC UNIVARIATE. Differences between LRFI and HRFI groups were analyzed using Student’s t-test for normally distributed data or the Mann–Whitney U test for non-normally distributed data. For microbiota analyses, alpha diversity indices were compared using the Wilcoxon rank-sum test, whereas beta diversity differences were assessed using permutational multivariate analysis of variance (PERMANOVA) based on Bray–Curtis dissimilarities with 999 permutations. LEfSe was used as an exploratory analysis with a nominal P value < 0.05 and an LDA score > 2.5; LEfSe P values were not adjusted for multiple testing. Genus-level differential abundance was further evaluated using the Wilcoxon rank-sum test followed by Benjamini–Hochberg correction, and FDR-adjusted P < 0.05 was considered significant. For metabolomics analyses, differential metabolites were identified based on VIP and fold-change thresholds combined with univariate statistical testing. Data are presented as mean ± SEM. Statistical significance was considered at P < 0.05, whereas 0.05 ≤ P < 0.10 was considered a statistical trend. Results Differences in production traits between LRFI and HRFI laying ducks As shown in Table 2, LRFI ducks exhibited a clear feed-efficiency advantage over HRFI ducks, characterized by significantly lower average daily feed intake (ADFI), residual feed intake (RFI), and feed conversion ratio (FCR) (all P < 0.05), whereas maintaining comparable body weight-related traits. No significant differences were observed between LRFI and HRFI ducks in initial body weight (BW), final BW, metabolic BW (MBW), body weight gain (BWG), or average daily gain (ADG) (all P > 0.05). In addition, LRFI ducks showed significantly higher average daily egg mass (ADEM) and egg production rate (EPR) than HRFI ducks (ADEM: 71.58 vs 62.85 g/d, P = 0.001; EPR: 99.9% vs 85.88%, P < 0.001; Table 2). Correlation analysis (Table 3, Spearman rank) further showed that RFI was strongly positively correlated with ADFI (ρ = 0.88, P < 0.01) and FCR (ρ = 0.85, P < 0.01), but negatively correlated with ADEM (ρ = −0.43, P < 0.01) and EPR (ρ = −0.49, P < 0.01). In addition, EPR was strongly positively correlated with ADEM (ρ = 0.71, P < 0.01).Table 2Effects of divergent residual feed intake (RFI) on growth performance and laying traits in Jinding ducks during 29 to 37 weeks of age (56 d).ItemsLRFIHRFISEMP-valueInitial BW (g)1507.91525.246.850.715Final BW (g)1655.81708.446.210.263MBW (g)250.7254.85.210.44BWG (g)147.9183.230.520.255ADG (g/d)2.643.270.550.255ADFI (g/d)160.2215.02.63< 0.001ADEM (g/d)71.5862.852.290.001EPR (%)99.9085.883.34< 0.001FCR (g/g)2.243.390.09< 0.001RFI (g/d)-38.4733.861.43< 0.001Abbreviations: BW = body weight; MBW = metabolic body weight; BWG = body weight gain; ADG = average daily gain; ADFI = average daily feed intake; ADEM = average daily egg mass; EPR = egg production rate; FCR = feed conversion ratio; RFI = residual feed intake. LRFI = low residual feed intake group; HRFI = high residual feed intake group. n = 20 ducks per group; residual degrees of freedom = 38. P-values were obtained using an independent-samples Student’s t-test. P < 0.05 was considered statistically significant.Table 3Spearman rank correlation coefficients (ρ) among residual feed intake (RFI), growth performance, feed efficiency, and laying traits in Jinding ducks.ItemInitial BWFinal BWMBWADGADFIADEMEPRFCRRFIInitial BW1Final BW0.73⁎⁎1MBW0.91⁎⁎0.94⁎⁎1ADG-0.300.34*0.041ADFI0.080.38*0.250.48*1ADEM0.130.130.140.01-0.201EPR-0.07-0.14-0.12-0.15-0.40*0.71⁎⁎1FCR-0.260.140.050.270.76⁎⁎-0.70⁎⁎-0.67⁎⁎1RFI-0.040.140.040.35*0.88⁎⁎-0.43⁎⁎-0.49⁎⁎0.85⁎⁎1Abbreviations: BW = body weight; MBW = metabolic body weight; ADG = average daily gain; ADFI = average daily feed intake; ADEM = average daily egg mass; EPR = egg production rate; FCR = feed conversion ratio; RFI = residual feed intake. Spearman rank correlation coefficients were calculated using individual birds as experimental units (n = 20 per group). P-values were two-tailed.⁎P < 0.05,⁎⁎P < 0.01. Table 2Effects of divergent residual feed intake (RFI) on growth performance and laying traits in Jinding ducks during 29 to 37 weeks of age (56 d).ItemsLRFIHRFISEMP-valueInitial BW (g)1507.91525.246.850.715Final BW (g)1655.81708.446.210.263MBW (g)250.7254.85.210.44BWG (g)147.9183.230.520.255ADG (g/d)2.643.270.550.255ADFI (g/d)160.2215.02.63< 0.001ADEM (g/d)71.5862.852.290.001EPR (%)99.9085.883.34< 0.001FCR (g/g)2.243.390.09< 0.001RFI (g/d)-38.4733.861.43< 0.001Abbreviations: BW = body weight; MBW = metabolic body weight; BWG = body weight gain; ADG = average daily gain; ADFI = average daily feed intake; ADEM = average daily egg mass; EPR = egg production rate; FCR = feed conversion ratio; RFI = residual feed intake. LRFI = low residual feed intake group; HRFI = high residual feed intake group. n = 20 ducks per group; residual degrees of freedom = 38. P-values were obtained using an independent-samples Student’s t-test. P < 0.05 was considered statistically significant. Effects of divergent residual feed intake (RFI) on growth performance and laying traits in Jinding ducks during 29 to 37 weeks of age (56 d). Abbreviations: BW = body weight; MBW = metabolic body weight; BWG = body weight gain; ADG = average daily gain; ADFI = average daily feed intake; ADEM = average daily egg mass; EPR = egg production rate; FCR = feed conversion ratio; RFI = residual feed intake. LRFI = low residual feed intake group; HRFI = high residual feed intake group. n = 20 ducks per group; residual degrees of freedom = 38. P-values were obtained using an independent-samples Student’s t-test. P < 0.05 was considered statistically significant. Table 3Spearman rank correlation coefficients (ρ) among residual feed intake (RFI), growth performance, feed efficiency, and laying traits in Jinding ducks.ItemInitial BWFinal BWMBWADGADFIADEMEPRFCRRFIInitial BW1Final BW0.73⁎⁎1MBW0.91⁎⁎0.94⁎⁎1ADG-0.300.34*0.041ADFI0.080.38*0.250.48*1ADEM0.130.130.140.01-0.201EPR-0.07-0.14-0.12-0.15-0.40*0.71⁎⁎1FCR-0.260.140.050.270.76⁎⁎-0.70⁎⁎-0.67⁎⁎1RFI-0.040.140.040.35*0.88⁎⁎-0.43⁎⁎-0.49⁎⁎0.85⁎⁎1Abbreviations: BW = body weight; MBW = metabolic body weight; ADG = average daily gain; ADFI = average daily feed intake; ADEM = average daily egg mass; EPR = egg production rate; FCR = feed conversion ratio; RFI = residual feed intake. Spearman rank correlation coefficients were calculated using individual birds as experimental units (n = 20 per group). P-values were two-tailed.⁎P < 0.05,⁎⁎P < 0.01. Spearman rank correlation coefficients (ρ) among residual feed intake (RFI), growth performance, feed efficiency, and laying traits in Jinding ducks. Abbreviations: BW = body weight; MBW = metabolic body weight; ADG = average daily gain; ADFI = average daily feed intake; ADEM = average daily egg mass; EPR = egg production rate; FCR = feed conversion ratio; RFI = residual feed intake. Spearman rank correlation coefficients were calculated using individual birds as experimental units (n = 20 per group). P-values were two-tailed. P < 0.05, P < 0.01. Cecal microbiota composition differs between LRFI and HRFI ducks A total of 3,778,295 input sequences were obtained from the 40 cecal samples. After quality filtering, 3,587,500 sequences were retained, corresponding to an overall quality-filtering retention rate of 94.95%. Following denoising, paired-end merging, and chimera removal, 1,840,108 non-chimeric sequences remained. The number of non-chimeric sequences ranged from 24,384 to 69,081 per sample, with an overall final retention rate of 48.70% relative to the input sequences (Supplementary Table S1). After singleton-ASV removal, the number of retained sequences ranged from 23,602 to 66,486 per sample. The number of observed ASVs after singleton removal and before rarefaction ranged from 3,747 to 10,573 per sample, with a median of 6,746. The singleton-filtered ASV table was subsequently rarefied to 22,422 sequences per sample, corresponding to approximately 95% of the minimum post-filtering library size, and all 40 samples were retained for diversity analyses. Rarefaction curves approached saturation, supporting adequate sequencing depth for downstream microbial diversity analyses (Supplementary Figure S1). To determine whether feed-efficiency divergence was accompanied by alterations in the cecal microbial ecosystem, we profiled the cecal microbiota of LRFI and HRFI ducks by 16S rRNA gene sequencing (n = 20 per group; Supplementary Table S2). Alpha diversity, assessed using Chao1, Shannon, Simpson, and Pielou’s evenness indices, was significantly higher in HRFI ducks than in LRFI ducks (Chao1 and Shannon, both P < 0.001; Simpson, P = 0.040; Pielou, P = 0.004; Fig. 1A), indicating increased richness, diversity, and evenness in the HRFI group. Bray–Curtis PCoA showed a partial separation between LRFI and HRFI ducks (PCoA1 = 11.6%, PCoA2 = 7.4%), and PERMANOVA confirmed a significant difference in community composition (P = 0.001, R² = 0.111; Fig. 1B). At the phylum level, the cecal microbiota was dominated by Bacteroidota and Bacillota, followed by Thermodesulfobacteriota, Pseudomonadota, and Actinomycetota (Fig. 1C). At the genus level, the predominant taxa included Bacteroides, Phocaeicola, Phascolarctobacterium, Blautia, and Desulfomicrobium (Fig. 1D).Fig. 1Cecal microbiota differences between low- and high-residual feed intake (RFI) ducks. (A) α-diversity indices (Chao1, Shannon, Simpson, Pielou). (B) Bray–Curtis PCoA. (C–D) Taxonomic composition at phylum (C) and genus (D) levels. (E) LEfSe-discriminative taxa (LDA > 2.5). (F) Relative abundance of seven representative differential genera (Wilcoxon rank-sum with Benjamini–Hochberg correction). LRFI = low residual feed intake group (higher feed efficiency); HRFI = high residual feed intake group (lower feed efficiency). PCoA = principal coordinate analysis; LEfSe = linear discriminant analysis effect size; LDA = linear discriminant analysis; FDR = false discovery rate. * FDR-adjusted P < 0.05; ** FDR-adjusted P < 0.01; All analyses included 20 ducks per group. Alpha-diversity indices were compared using the Wilcoxon rank-sum test, beta diversity was assessed using PERMANOVA with 999 permutations, and genus-level differential abundance was evaluated using the Wilcoxon rank-sum test followed by Benjamini–Hochberg correction. LEfSe was considered exploratory and used a nominal P value < 0.05 and an LDA score > 2.5 as the selection criteria. Cecal microbiota differences between low- and high-residual feed intake (RFI) ducks. (A) α-diversity indices (Chao1, Shannon, Simpson, Pielou). (B) Bray–Curtis PCoA. (C–D) Taxonomic composition at phylum (C) and genus (D) levels. (E) LEfSe-discriminative taxa (LDA > 2.5). (F) Relative abundance of seven representative differential genera (Wilcoxon rank-sum with Benjamini–Hochberg correction). LRFI = low residual feed intake group (higher feed efficiency); HRFI = high residual feed intake group (lower feed efficiency). PCoA = principal coordinate analysis; LEfSe = linear discriminant analysis effect size; LDA = linear discriminant analysis; FDR = false discovery rate. * FDR-adjusted P < 0.05; ** FDR-adjusted P < 0.01; All analyses included 20 ducks per group. Alpha-diversity indices were compared using the Wilcoxon rank-sum test, beta diversity was assessed using PERMANOVA with 999 permutations, and genus-level differential abundance was evaluated using the Wilcoxon rank-sum test followed by Benjamini–Hochberg correction. LEfSe was considered exploratory and used a nominal P value < 0.05 and an LDA score > 2.5 as the selection criteria. LEfSe was used as an exploratory approach to identify discriminative taxa between groups using a nominal P value < 0.05 and an LDA score > 2.5 (Fig. 1E; Supplementary Table S3). This analysis indicated that Phocaeicola, Desulfomicrobium, Saccharopolyspora, Cytophaga, Porphyromonas, Bradyrhizobium, Olsenella, and Marivirga were enriched in LRFI ducks, whereas Bacteroides, Blautia, Lucifera, Collinsella, Hungatella, and Mitsuokella were enriched in HRFI ducks. To further assess the robustness of these differences, Wilcoxon rank-sum tests followed by Benjamini–Hochberg correction identified 15 genera with significant between-group differences (FDR-adjusted P < 0.05; Supplementary Table S4), among which seven representative genera are shown in Fig. 1F. Notably, the direction of enrichment for these representative genera was consistent with the LEfSe results. Specifically, Phocaeicola, Bradyrhizobium, and Olsenella were more abundant in LRFI ducks, whereas Bacteroides, Collinsella, Flavonifractor, and Eggerthella were enriched in HRFI ducks (Fig. 1F). Cecal metabolomic profiles differ between LRFI and HRFI ducks Untargeted metabolomics of cecal contents further revealed marked differences in cecal metabolomic profiles between LRFI and HRFI ducks. (n = 20 per group; Fig. 2). Following peak detection, feature alignment, and quality-control filtering, 1,828 ion features in ESI+ mode and 1,094 ion features in ESI− mode were retained for statistical analysis before metabolite annotation. QC samples clustered tightly in PCA space in both ESI+ and ESI− modes (Supplementary Figure S2), indicating stable analytical performance and good repeatability throughout the run. Unsupervised PCA showed a tendency toward group separation between LRFI and HRFI samples in both ESI+ and ESI− modes (Fig. 2A and Supplementary Figure S3A). Supervised OPLS-DA further improved group discrimination (Fig. 2B and Supplementary Figure S3B). The OPLS-DA models showed good goodness-of-fit (R²Y = 0.888 in ESI+; R²Y = 0.862 in ESI−), with moderate predictive ability in ESI+ (Q² = 0.468) and relatively weak predictive ability in ESI− (Q² = 0.244). Permutation tests (200 permutations) suggested no obvious overfitting in either ionization mode. In ESI+ mode, the Q² intercept was −0.17 (Fig. 2C), whereas in ESI− mode the Q² intercept was −0.19 (R² intercept = 0.79; Supplementary Figure S3C). In both cases, the permuted R²/Q² values were generally lower than those of the original models, indicating a low risk of overfitting.Fig. 2Untargeted metabolomics reveals distinct cecal metabolic signatures between LRFI and HRFI laying ducks. (A) PCA score plot (ESI+). (B) OPLS-DA score plot (ESI+). (C) Permutation test for the OPLS-DA model (200 permutations, ESI+). (D) Volcano plot of differential metabolites (x: log2FC; y: −log10 P); the full list is provided in Supplementary Table S5. (E) KEGG-based pathway mapping of differential metabolites (ranked by nominal P). (F,G) Relative abundance of representative differential metabolites between different groups analyzed using the Wilcoxon rank-sum test. LRFI = low residual feed intake group (higher feed efficiency); HRFI = high residual feed intake group (lower feed efficiency). PCA = principal component analysis; OPLS-DA = orthogonal partial least squares discriminant analysis; ESI+ = positive electrospray ionization mode; FC = fold change; KEGG = Kyoto Encyclopedia of Genes and Genomes. * P < 0.05, ** P < 0.01, ***P < 0.001; n = 20 per group. Untargeted metabolomics reveals distinct cecal metabolic signatures between LRFI and HRFI laying ducks. (A) PCA score plot (ESI+). (B) OPLS-DA score plot (ESI+). (C) Permutation test for the OPLS-DA model (200 permutations, ESI+). (D) Volcano plot of differential metabolites (x: log2FC; y: −log10 P); the full list is provided in Supplementary Table S5. (E) KEGG-based pathway mapping of differential metabolites (ranked by nominal P). (F,G) Relative abundance of representative differential metabolites between different groups analyzed using the Wilcoxon rank-sum test. LRFI = low residual feed intake group (higher feed efficiency); HRFI = high residual feed intake group (lower feed efficiency). PCA = principal component analysis; OPLS-DA = orthogonal partial least squares discriminant analysis; ESI+ = positive electrospray ionization mode; FC = fold change; KEGG = Kyoto Encyclopedia of Genes and Genomes. * P < 0.05, ** P < 0.01, ***P < 0.001; n = 20 per group. Using the criteria of VIP > 1, FC > 1.5 or FC < 0.67, and raw P < 0.05, 251 annotated differential features were identified, including 148 features that were increased and 103 that were decreased in LRFI ducks relative to HRFI ducks (Fig. 2D). Of these, 77 were classified as Level 1 compounds confidently identified by matching against the authentic-standard-based in-house reference library, 129 as Level 2 putatively annotated compounds, and 45 as Level 3 tentatively annotated compounds according to the annotation-confidence criteria provided by the analytical service provider (Supplementary Table S5). To place these metabolic shifts into a functional context, KEGG-based pathway mapping was performed for the differential features (Fig. 2E). The enrichment bubble plot indicated that these metabolites were mainly assigned to pathways related to energy metabolism, amino acid metabolism, and lipid-associated processes. However, no KEGG pathway passed multiple-testing correction. Therefore, pathway annotations are presented descriptively to summarize recurring functional themes, with lipid handling and bile acid-related metabolism among the top-ranked categories. Representative differential metabolites were selected and visualized as boxplots according to their direction of change (Fig. 2F–G). Metabolites elevated in LRFI ducks (Fig. 2F) were associated with diverse biochemical classes, including lipid-related molecules (e.g., phospholipids), fat-soluble vitamin derivatives, and plant-derived metabolites, whereas metabolites reduced in LRFI ducks (i.e., relatively higher in HRFI; Fig. 2G) were more often associated with xenobiotic- and aromatic compound-related metabolism. Together, these results suggest that feed-efficiency divergence was accompanied by broad metabolic differences in the cecal environment. Lipid-, energy-, and amino acid-related categories appeared among the top-ranked pathways, but these patterns should be considered descriptive and hypothesis-generating. Integrated correlation analysis reveals coordinated features associated with the LRFI–HRFI phenotypic contrast To examine whether differential microbial and metabolic features showed coordinated relationships with feed-efficiency-related phenotypes, Spearman correlation analyses were performed across the microbial, metabolic, and phenotypic domains. Several genera enriched in the LRFI group, including Phocaeicola, Olsenella, and Bradyrhizobium, were negatively correlated with ADFI, FCR, and RFI but positively correlated with EPR. In contrast, genera enriched in the HRFI group, such as Bacteroides, Collinsella, and Niallia, exhibited the opposite pattern, being positively associated with ADFI, FCR, and RFI and negatively associated with EPR (Fig. 3A). Overall, these genus–trait correlations revealed that LRFI-associated taxa tended to correlate with an LRFI-associated phenotypic profile, characterized by lower ADFI, FCR, and RFI together with higher EPR.Fig. 3Integrated correlations among cecal microbiota, metabolites, and feed-efficiency traits in laying ducks. (A) Spearman correlations between differential genera (16S) and phenotypic traits (RFI, FCR, ADFI, EPR). (B) Spearman correlations between RFI-correlated differential metabolites and the same traits. (C) Genus–metabolite correlation heatmap highlighting module-like relationships; only |ρ| > 0.5 and P < 0.05 are shown. (D) Mantel test–based network showing correspondence between phenotypic dissimilarity and integrated omics dissimilarity across individuals (Bray–Curtis; Spearman). (E) Boxplots of representative genera embedded in the phenotype–metabolite correlation network. (F) Relative abundance of representative metabolites embedded in the same network. Only associations meeting |ρ| ≥ 0.5 and nominal P < 0.05 are displayed to reduce visual complexity. For the correlation analyses, P values were nominal and were not adjusted for multiple testing. All analyses included 20 ducks per group. LRFI = low residual feed intake group (higher feed efficiency); HRFI = high residual feed intake group (lower feed efficiency). RFI = residual feed intake; FCR = feed conversion ratio; ADFI = average daily feed intake; EPR = egg production rate. * P < 0.05, ** P < 0.01, *** P < 0.001. Integrated correlations among cecal microbiota, metabolites, and feed-efficiency traits in laying ducks. (A) Spearman correlations between differential genera (16S) and phenotypic traits (RFI, FCR, ADFI, EPR). (B) Spearman correlations between RFI-correlated differential metabolites and the same traits. (C) Genus–metabolite correlation heatmap highlighting module-like relationships; only |ρ| > 0.5 and P < 0.05 are shown. (D) Mantel test–based network showing correspondence between phenotypic dissimilarity and integrated omics dissimilarity across individuals (Bray–Curtis; Spearman). (E) Boxplots of representative genera embedded in the phenotype–metabolite correlation network. (F) Relative abundance of representative metabolites embedded in the same network. Only associations meeting |ρ| ≥ 0.5 and nominal P < 0.05 are displayed to reduce visual complexity. For the correlation analyses, P values were nominal and were not adjusted for multiple testing. All analyses included 20 ducks per group. LRFI = low residual feed intake group (higher feed efficiency); HRFI = high residual feed intake group (lower feed efficiency). RFI = residual feed intake; FCR = feed conversion ratio; ADFI = average daily feed intake; EPR = egg production rate. * P < 0.05, ** P < 0.01, *** P < 0.001. Among the differential metabolites, the subset displayed in Fig. 3B also largely followed an LRFI-associated correlation profile, showing negative associations with ADFI, FCR, and RFI and positive associations with EPR. Among the differential metabolites, 21 met the exploratory correlation criteria of |ρ| ≥ 0.5 and nominal P < 0.05 in relation to RFI, of which 15 representative metabolites are shown. Notably, most highlighted metabolites, including omega-Hydroxyphylloquinone and Ginkgolide C, were negatively associated with ADFI, FCR, and RFI but positively associated with EPR, consistent with an LRFI-associated profile. In contrast, PC (P-16:0/24:0) and omega-Hydroxy-beta-dihydromenaquinone-9 showed the opposite direction, exhibiting positive associations with ADFI, FCR, and RFI and negative associations with EPR, in line with the HRFI-associated phenotype (Fig. 3B; full list in Supplementary Table S6). Integrated genus–metabolite correlation analysis further revealed coordinated correlation patterns between differential genera and metabolites (Fig. 3C). Key LRFI-enriched genera, including Olsenella and Phocaeicola, were positively correlated with multiple LRFI-associated metabolites but negatively correlated with HRFI-associated metabolites. Conversely, HRFI-enriched genera, including Collinsella, Eggerthella, and Niallia, showed reciprocal correlation patterns. At the system level, Mantel test–based analysis indicated significant associations between feed-efficiency-related phenotypic variation and combined microbial–metabolic variation across individuals. Network visualization further linked these phenotypic traits with selected microbial and metabolic features (Fig. 3D). Consistent with this pattern, representative genera jointly associated with phenotypic traits and metabolites, such as Phocaeicola and Olsenella, differed significantly between LRFI and HRFI ducks (Fig. 3E), and representative metabolites embedded in the same correlation network, including omega-Hydroxyphylloquinone and Ginkgolide C, also differed between groups (Fig. 3F). Collectively, these analyses revealed coordinated cross-domain relationships among differential microbial genera, metabolites, and feed-efficiency-related traits and highlighted microbial–metabolic feature combinations associated with RFI divergence. Discussion Feed efficiency is a key production trait in laying ducks, but the biological basis of individual variation remains incompletely understood. In the present study, feed-efficiency divergence in laying Jinding ducks was clearly reflected at the phenotypic level and occurred alongside distinct differences in cecal microbiota and metabolomic profiles. Although RFI is generally considered less dependent on production traits than FCR, LRFI ducks in the present study also exhibited higher EPR and ADEM than HRFI ducks. Therefore, the microbial and metabolic differences observed between groups cannot be attributed solely to RFI and may also reflect the concurrent differences in egg production rate and average daily egg mass. LRFI ducks consumed substantially less feed than HRFI ducks, with ADFI reduced from 215.0 to 160.2 g/d, whereas body weight-related traits remained similar between groups. Accordingly, LRFI ducks showed lower RFI and FCR, indicating better feed efficiency under the present conditions. These findings are consistent with the interpretation of RFI as an intake-based indicator of feed efficiency that is less influenced by production level or body size (Zeng et al., 2017; Chang et al., 2024; Liu et al., 2024; Wan et al., 2025). Beyond phenotypic differences, feed-efficiency divergence was also associated with distinct shifts in cecal microbial ecology. HRFI ducks showed higher richness, diversity, and evenness than LRFI ducks, together with a clear difference in beta diversity. Higher alpha diversity, however, does not necessarily indicate a more favorable or more efficient microbiota. Previous studies have suggested that diversity can also reflect differences in substrate availability and ecological niche space rather than efficiency itself (Shade, 2017; Saup et al., 2019; Shi et al., 2025). In the present study, HRFI ducks had higher ADFI than LRFI ducks, which may have increased the amount of residual substrate entering the cecum and supported a broader range of taxa (Musigwa et al., 2021; Elling-Staats et al., 2022). Similar intake- or substrate-related shifts in gut microbiota have also been reported in poultry under different feeding conditions (Metzler-Zebeli et al., 2019). Together, these results suggest that diversity differences in this study were more likely related to substrate flow and community composition and should not be interpreted simply as indicating a more or less “efficient” microbiota. In addition to diversity, several differential genera showed consistent associations with feed-efficiency divergence. LEfSe and Wilcoxon-BH analyses both identified LRFI-enriched genera such as Phocaeicola, Olsenella, and Bradyrhizobium, whereas HRFI ducks were enriched in Bacteroides, Collinsella, Flavonifractor, and Eggerthella. Among the genera enriched in LRFI ducks, Phocaeicola and Olsenella are of particular biological interest because of their reported carbohydrate-fermenting capacities. Members of Phocaeicola, including P. vulgatus, can utilize diverse carbohydrates through fermentative metabolic pathways (Clausen et al., 2024). Cultured members of Olsenella have been characterized as anaerobic carbohydrate-fermenting bacteria that produce lactate and acetate (Kraatz et al., 2011; Paek et al., 2024). These reported characteristics suggest that the enrichment of these genera may be associated with differences in cecal carbohydrate fermentation and the local organic-acid environment between LRFI and HRFI ducks. Previous poultry studies have similarly linked cecal microbial fermentation and microbiota composition with variation in feed efficiency (Yan et al., 2017; Wen et al., 2021). The observed correlations of Phocaeicola and Olsenella with ADFI, FCR, RFI, and EPR were consistent with the broader LRFI–HRFI phenotypic contrast, but do not demonstrate their metabolic activity or an independent effect on feed efficiency. Direct quantification of cecal fermentation products, together with species- or strain-resolved functional analyses, will be required to test their proposed roles. Consistent with the microbial differences, metabolomic profiling revealed distinct differences between LRFI and HRFI ducks, with lipid-, bile acid-, energy-, and amino acid-related processes appearing among the top-ranked functional themes (Patience et al., 2015; Abd El-Hack et al., 2025). However, no KEGG pathway remained significant after multiple-testing correction. Therefore, these pathway-level patterns should be regarded as exploratory observations rather than statistically established metabolic alterations. Metabolite annotation confidence varied among compounds. Level 1 metabolites were confidently identified by matching against an in-house reference library established using authentic standards, whereas Level 2 and Level 3 annotations were based primarily on public database or spectral-library evidence. Because database-based annotations may not unambiguously distinguish structural isomers or compounds with similar fragmentation patterns, interpretations of Level 2 and Level 3 metabolites and their associated pathways should remain tentative. Lipid-related metabolism may nevertheless warrant further investigation in laying birds. Unlike broilers, in which feed efficiency is often discussed mainly in relation to growth and fat deposition, laying ducks must continuously support hepatic lipogenesis and lipid transport for yolk formation during the production period. Several individual differential metabolites were putatively annotated as lipid-related compounds, which may be relevant to the specific metabolic demands of egg production under divergent feed-efficiency phenotypes (Guo et al., 2025; Zhong et al., 2025). These observations provide a rationale for future targeted lipidomic and bile acid analyses. The integrated analysis further connected the microbial and metabolic differences with feed-efficiency-related phenotypes. LRFI- and HRFI-associated genera showed broadly concordant relationships with their corresponding metabolic and phenotypic profiles. The concordant directions across microbial, metabolic, and phenotypic domains suggest a coordinated cecal ecological and metabolic pattern accompanying RFI divergence, rather than isolated differences in individual taxa or metabolites. These cross-domain relationships highlight microbial–metabolic combinations for subsequent targeted investigation. However, because the associations were derived from pairwise correlations within the same extreme-phenotype population, multivariable assessment and validation in independent populations will be needed to determine their independent and predictive value. Previous RFI studies in ducks provide important context for the present findings. Consistent with earlier reports, LRFI ducks in this study consumed less feed and exhibited a lower FCR than HRFI ducks (Zeng et al., 2018; Sun et al., 2023; Chang et al., 2024). The LRFI ducks selected for the omics analyses also showed higher EPR and ADEM. This difference should be interpreted in the context of the extreme-phenotype sampling design, because the omics subset comprised the 20 most extreme individuals selected from each 50-duck divergent RFI pool and was not matched for laying performance. Thus, the stronger production-trait contrast observed in the selected 40 ducks may partly reflect the phenotypic structure generated by extreme-tail selection. Accordingly, the microbial and metabolic differences identified in this subset characterize the combined LRFI–HRFI phenotypic contrast and cannot be attributed specifically to RFI or interpreted as evidence that low RFI directly improves reproductive performance. Taxon-level findings have not been fully consistent across duck RFI studies, potentially because of differences in breed, laying stage, diet, intestinal sampling site, analytical procedures, phenotyped population size, and selection intensity. The present study extends previous work by obtaining matched cecal microbiota and untargeted metabolomic profiles from the same 40 extreme-RFI Jinding ducks selected from a population of 767 phenotyped birds. This design enabled individual-level assessment of relationships among microbial taxa, metabolites, feed-efficiency traits, and laying-performance traits. Several limitations should be acknowledged. This observational comparison cannot establish causality or directionality. The moderate sample size (n = 20 per group) was powered mainly for relatively large effects, so smaller differences may have been missed. Additionally, 16S rRNA sequencing provides genus-level profiles rather than strain-resolved functions, and many metabolite annotations remain putative pending authentic-standard confirmation. Future strain-resolved and targeted quantitative analyses of bile acids, SCFAs, and lipid mediators will help clarify the functional links between gut microbiota and feed efficiency in laying ducks. Conclusion In conclusion, divergent residual feed intake (RFI) phenotypes in laying Jinding ducks were characterized by lower feed intake and improved feed efficiency without significant differences in body weight-related traits. Integrated cecal 16S rRNA gene sequencing and untargeted metabolomics revealed clear differences in microbiota composition and metabolomic profiles between LRFI and HRFI ducks. Exploratory KEGG analysis highlighted several top-ranked metabolic categories, but no pathway remained significant after multiple-testing correction. Correlation analyses further indicated coordinated associations among differential genera, metabolites, and feed-efficiency-related traits. These findings suggest that RFI divergence in laying ducks is associated with distinct cecal microbiota and metabolomic features. The taxa and metabolite classes identified here represent candidate features associated with the LRFI–HRFI phenotypic contrast. Further experimental and independent-cohort studies are needed before their functional relevance or potential value as intervention targets can be established. Ethics approval All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Shihezi University, Shihezi, China (Authorization No. A2025-545), and were conducted in accordance with the institutional guidelines for the care and use of experimental animals. As the primary institution responsible for this study, Jiangsu Agri-animal Husbandry Vocational College supervised the research and ensures adherence to these ethical standards. Data and model availability statement The sequencing data have been deposited in the NCBI SRA under accession number PRJNA1418964 and will be publicly available upon publication. All other data generated or analyzed during this study are included in this published article and its supplementary information files. Further inquiries can be directed to the corresponding author. Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work, the authors used ChatGPT in order to improve language and readability. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. AI-assisted technologies were not used for data analysis, interpretation, or scientific decision-making. Financial support statement This work was supported by the National Natural Science Foundation of China (Grant No. 31660703), the Key Technology Integration and Demonstration Promotion for Intensive Waterfowl Farming (JCTG[2025]16), and Taizhou Science and Technology Support Program (Social Development) Project (TS202428). Disclosures The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.