A transcribed, multi-channel, and continuously evolving molecular recorder 77 To achieve our goal of a tunable, high information content molecular recorder, we 78 utilized Cas9 to generate insertions or deletions (indels) upon repair of double-stranded breaks, 79 which are inherited in the next generation of cells 11-16. We record within a 205 base pair, synthetic DNA "target site" containing three "cut sites" and a static 8 base pair "integration barcode" (intBC), which are delivered in multiple copies via piggyBac transposition (Fig. 1a, b). We embedded this sequence into the 3'UTR of a constitutively transcribed fluorescent protein to enable profiling from the transcriptome. A second cassette encodes three independently transcribed and complementary guide RNAs to permit recording of multiple, distinct signals (Fig. 1a, b) 18. Our system is capable of high information storage due to the diversity of heritable repair outcomes, and the large number of targeted sites, which can be distinguished by the intBC (Fig. 1c). DNA repair generates hundreds of unique indels, and the distribution for each cut site is different and nonuniform: some produce highly biased outcomes while others create a diverse series (Fig. 1c, Extended Data Fig. 1) 19-21. To identify sequences that can tune the mutation rate of our recorder for timescales that are not pre-defined, and may extend from days to months, we screened several guide RNA series containing mismatches to their targets 22 by monitoring their activity on a GFP reporter over a 20-day timecourse and selected those that demonstrated a broad dynamic range (Fig. 1d). Slower cutting rates may improve viability in vivo, as frequent Cas9mediated double-strand breaks can cause cellular toxicity 23,24. To demonstrate information recovery from single cell transcriptomes, we stably transduced K562 cells with our technology and generated a primary, cell-barcoded cDNA pool via the 10x Genomics platform, allowing us * * *

We present Vision, a tool for annotating the sources of variation in single cell RNA-seq data in an automated and scalable manner. Vision operates directly on the manifold of cell-cell similarity and employs a flexible annotation approach that can operate either with or without preconceived stratification of the cells into groups or along a continuum. We demonstrate the utility of Vision in several case studies and show that it can derive important sources of cellular variation and link them to experimental meta-data even with relatively homogeneous sets of cells. Vision produces an interactive, low latency and feature rich web-based report that can be easily shared among researchers, thus facilitating data dissemination and collaboration.

No abstract