Skip to main content
Premium Trial:

Request an Annual Quote

Nature Papers Describe Method to Predict Genome Folding, Comparison of Massively Parallel Reporter Assays

A computational method for predicting genome folding using only DNA sequence as an input is reported in Nature Methods this week. The convolutional neural network, dubbed Akita, takes in approximately one megabase of DNA sequence and predicts contact frequency maps for all pairs of around 2-kb bins within this region, according to the method's developers at the Gladstone Institute of Data Science and Biotechnology and Calico Life Sciences. The scientists demonstrate that Akita enables rapid in silico mutagenesis at the motif and single nucleotide level and how it can be used to interpret expression quantitative trait loci, and make predictions for multi-kilobase structural variants and probe species-specific genome folding.

A systematic comparison of several massively parallel reporter assays (MPRAs) is published in Nature Methods this week, showing that assay design can influence results. In the study, investigators from the University of Washington and their collaborators screen the same library of 2,440 candidate liver enhancers and controls for regulatory activity in HepG2 cells using nine different MPRA strategies including conventional episomal, self-transcribing active regulatory region sequencing, and lentiviral designs. Additionally, they tested the same sequences in both orientations. The researchers uncover "subtle but significant differences that correlate with epigenetic and sequence-level features, as well as differences in dynamic range and reproducibility," with sequence length having the greatest effect, followed by assay design and orientation. "This work provides a framework for the experimental design of high-throughput reporter assays, suggesting that the extended sequence context of tested elements and to a lesser degree the precise assay, influence MPRA results," the researchers write.