A new method for profiling genomic and epigenetic landscapes at the single-cell level is reported in Nature Biotechnology this week. To address the challenges facing high-throughput, single-cell assessment of heterochromatin and its underlying genomic determinants, a team led by scientists at the San Raffaele Scientific Institute in Italy developed a single-cell technique, based on the engineering of a Tn5 transposase targeting H3K9me3, that comprehensively probes both open and closed chromatin and concomitantly records the underlying genomic sequences. They tested the method — dubbed single-cell genome and epigenome by transposases sequencing, or scGET-seq — in cancer-derived organoids and human-derived xenograft models and identified genetic events and plasticity-driven mechanisms contributing to cancer drug resistance. Using the epigenetic data generated with scGET-seq, they then built a computational approach called Chromatin Velocity that defines vectors of cellular fate and predicts future cell states based on the ratio between open and closed chromatin.
A machine learning method that reconstructs full-length genotype frequencies, trajectories, and fitness from short-read sequencing data derived during directed evolution campaigns is presented in Nature Chemical Biology this week. Directed evolution can be used to generate proteins with novel activities. However, it remains difficult to measure full-length genotypes, their frequencies, and fitness for evolving gene-length biomolecules using most high-throughput DNA sequencing methods since short-read lengths can lose mutation linkages in haplotypes. To overcome this issue, a trio of Broad Institute scientists developed Evoracle, a computational approach that uses covariation in point mutation frequencies over time to accurately reconstruct the fitness and frequencies of full-length genotypes from directed evolution timepoints, even with very short DNA sequencing read lengths and substantial measurement noise such as sequencing errors. They show that Evoracle outperforms two related overlap-free gene reconstruction methods and retains strong performance with pooled Sanger sequencing data. They also show that it can propose variants with higher fitness than common approaches that use consensus mutations.