A significant challenge in the field of microbiology is the functional annotation of sequence novel genes from microbiomes. The increasing pace of sequencing technology development has made solving this challenge in a high-throughput manner even more important. Functional metagenomics offer a sequence-naïve and cultivation-independent solution. This forward genetics approach relies on the creation of functional metagenomic libraries (aka shotgun cloning) in which a microbial host such asE. coliis transformed with vectors containing metagenomic DNA fragments and, optimally, expresses any captured genes into a corresponding phenotype. These libraries can be screened or selected for a function of interest, such as antibiotic resistance, allowing the captured metagenomic DNA to be linked to a phenotype regardless of the sequence’s novelty. Unfortunately, most methods for constructing functional metagenomic libraries require large input masses of metagenomic DNA, putting many sample types off limits to this toolset. Here, we show that our recently developed functional metagenomic library preparation method, METa assembly, can be used to prepare useful libraries from much lower input DNA masses. Standard methods of functional metagenomic library preparation generally call for 5 μg to 60 μg of input metagenomic DNA. Here, we demonstrate that the threshold for input DNA mass can be lowered at least to 30.5 ng, a three-log decrease from prior art. These functional metagenomic libraries, prepared using between 30.5 ng and 100 ng of metagenomic DNA, nonetheless were sufficient to link three MFS efflux pumps to tetracycline resistance and capture two potential genes for degradation or resistance to the antidiabetic pharmaceutical acarbose. Our preparation of functional metagenomic libraries from aquatic samples and a model fecal swab demonstrate that METa assembly can be used to prepare functional metagenomic libraries from microbiomes that were previously incompatible with this approach. Functional metagenomic screens and selections are one of the few high-throughput methods that can link novel genes to functions and here we show that one of their significant drawbacks, a requirement for large amounts of metagenomic DNA, can now be overcome.