Only few genomic variants are reported as associated to feed efficiency. To empower our analysis, we used imputation of sequence variants in two Large White pig designs (1,943 pigs and 1,942 pigs, respectively) with records for feed efficiency, and abundances of faecal microbiota genera from 16S rRNA gene sequencing (600 and 1,400 pigs, respectively). Imputation was carried out jointly on the two designs with FImpute, in two steps. First, the 650 K genotypes of 128 founders (Affymetrix Axiom Porcine Array) of the two designs were imputed to animals of the full designs, that had been genotyped with 60K SNP chips. Then, sequence variants, called on 45 founders of the two designs using the nf-core/sarek workflow, were imputed to full designs, leading to a total of 17,033,057 SNP imputed (correlation between true and imputed genotypes per animal = 0.94±0.011). After quality control, association studies were carried out with GEMMA in each design separately, and meta-analyses were used to combine the corresponding outputs with the metal package in R. The simpleM approach was used (principal component analysis applied to the matrix of genotypes) to estimate the number of independent tests, leading to a genome-wide threshold of 7.22 for the -log10(P-values). In total, 26 QTL regions were significant, and 272 regions were considered as suggestive. An enrichment analysis applied to all annotated genes in these regions pointed out GO terms related to membrane compounds and functions. Focusing on the 10 QTL regions associated with the abundances of 9 microbiota genera, functional candidate genes were suggested, for instance previously reported in colorectal cancer in mice or human. Altogether, the approach combining sizable datasets, imputation of sequence variants and records of traits related to sub-functions of complex traits allowed to pinpoint new pathways and genes related to variation of feed efficiency in pigs.