Availability of palaeogenomic information depends on the survival of ancient DNA (aDNA) in human remains. Therefore, studies often include only a limited number of individuals or rely on pre-selected genomic sites that restrict further analysis and do not utilize full information potential of the samples despite destructive aDNA sampling. While this is changing with the availability of whole genome sequencing, aDNA analysis is still easily influenced by contamination, poor DNA preservation, sequencing errors and reference bias. Incorporating these uncertainties directly to the analysis through genotype likelihoods results in increased power and accuracy for population genetic inferences. Our suite of statistical tools, ATLAS, enables users to build complete customised analysis pipelines, takes into account diverse sources of error and accurately estimates genotype likelihoods and allele frequency spectra to be further used in explicit modelling. It also provides several standalone inference methods that - among others - include reference-free determination of genetic diversity within and between individuals and populations. These genetic measures can be, together with individual ancestral affinities, directly compared to archaeological distance measures and interpreted in relation to archaeologically associated cultural markers. We demonstrate the utility of these approaches on palaeogenomic data obtained from up to 9,000 years old samples from sites associated with Lepenski Vir culture, including the settlement at the eponymous site. Genetic and cultural affinities of individuals from Mesolithic, Transition and Neolithic periods provide insights into an active role of these fisher-hunter-gatherers during the Neolithisation of the area of Central Balkans and the structure of their sedentary society.