Cellular proliferation biases clonal lineage tracing and trajectory inference

Abstract

We identify a fundamental statistical phenomenon in single-cell time courses with clone-based lineage tracing. Through simple probabilistic arguments, we show how the relative growth rates of cells influence the probability that they will be sampled in clones observed across multiple time points. We support these arguments with a simple simulation study and a time-course of T-cell development, and we demonstrate that this bias can impact fate probability predictions from trajectory inference methods. Finally, we explore how to develop trajectory inference methods which are robust to this bias. In particular, we show how to extend LineageOT to use data from clones observed across multiple time points.

Time- and lineage-resolved transcriptional profiling uncovers gene expression programs and clonal relationships that underlie human T lineage specification

“…S6). CoSpar may be prone to sampling bias due to differences in cellular growth rates 62 . LineageOT does not use multi-time clone information directly (it only uses clonal relationships within time-points) and it relies on estimates of cell growth rates.…”

“…We therefore utilized experimental measurements of cell growth rates (see methods), and used a modified version of LineageOT that incorporate multi-time data - a method called Multi-Time LOT (MT-LOT) (see methods and Bonham-Carter et al 62 ). The results from MT-LOT reflect a midpoint between the previous predictions from CoSpar and LOT and we proceeded using this experimentally-supported inference method.…”

“…We use the data obtained from the lineage barcoding experiment to generate transition maps using lineageOT with multi-time information for the time pairs (4, 5) and (9, 10) 62 . The following filter was applied for each time pair as the preprocessing step.…”