Extended Data Fig. 2: Dual color measurements to validate single-cell deconvolution and measure elongation rate.

(a) Validation of the kernel assumptions underlying the deconvolution of initiation events from single-allele transcription time series, using dual-color confocal imaging of hb and Kr. For hb (Kr), fly lines were generated with MS2 (PP7) stem-loop cassettes in the first intron and PP7 (MS2) cassettes in the 3′UTR. Both lines were visualized using MCP-GFP (green) and PCP-mCherry (red). As the two fluorescent signals are correlated via the elongation process, the paired time series impose stricter constraints on the underlying initiation events, making them an effective test of the deconvolution method. Deconvolution is performed jointly on each channel (that is, a single train of polymerases must explain both signals) using two kernels tailored to the stem-loop positions and satisfying the key assumptions: (i) constant, deterministic elongation rate; (ii) no Pol II pausing or dropping in the gene body; (iii) absence of co-transcriptional splicing; and (iv) fast termination. Additionally, the dual-color configuration allows estimation of the average elongation rate based on the time delay between the two signals and the known genomic distance between insertion sites. (b) Dual-color signal reconstruction from deconvolved single-allele time series (black lines: raw data). The transcription rate (gray line with ±1 s.d. envelope) is inferred from a single allele’s measured green and red traces. The denoised reconstructed signals (green and red, with ±1 s.d.) are obtained by re-convolving the inferred transcription rate with the appropriate kernel per channel. The close agreement between reconstructed and raw signals supports the validity of the kernel assumptions (see c). (c) Distribution of residuals between measured and reconstructed dual-color signals. Normalized residuals were computed as the difference between raw and reconstructed signals (panel b), divided by the standard deviation of imaging noise. This was done per allele for hb (N = 2666, blue) and Kr (N = 2594, pink). The spread of mean and standard deviation values of these residuals (black ellipse, 95% confidence) closely matches the expected distribution for a perfect model (dotted ellipse, 95% confidence), indicating good reconstruction fidelity. (d–e) Estimated elongation rate K_elo from dual-color measurements. (d) Mean elongation rate as a function of AP position, averaged over nuclei from 10 embryos, for both hb (blue) and Kr (pink) in NC13 (squares) and NC14 (circles). Error bars indicate the standard deviation across embryo means. (e) Per-embryo elongation rates (symbols and color scheme as in d), with error bars showing standard deviation across AP positions. Elongation rate is consistent across genes and nuclear cycles, with a global estimate of K_elo = 1.8 ± 0.1 kb/min (black line ±1 s.d., dashed).