Since the majority of microbial organisms still evade cultivation attempts, genomic insights into many taxa are limited to cultivation-independent approaches. However, current methods of metagenomics and single-cell genome sequencing have individual drawbacks, which can limit the quality as well as completeness of the reconstructed genomes. Current attempts to combine both approaches still use whole genome amplification techniques which are prone to bias. Here, we propose a novel approach for the purpose of genome reconstructions that utilizes the potential of cell sorting for targeted enrichment and depletion of different cell types to create distinct cell fractions with sufficient DNA amounts, circumventing amplification. By distributing sequencing efforts over these fractions as well as the original sample, coassemblies become highly optimized for coabundance variation based binning approaches. "Midi-metagenomics" enables accurate metagenome-assembled genome (MAG) reconstruction from individual sorted samples with higher quality than coassembly and binning of multiple distinct samples and therefore improves analyses of the so-called "microbial dark matter".