DNA Replication Timing Data Corroborate In Silico Human Replication Origin Predictions

We develop a wavelet-based multiscale pattern recognition methodology to disentangle the replication- from the transcription-associated compositional strand asymmetries observed in the human genome. Comparing replication skew profiles to recent high-resolution replication timing data reveals that most of the putative replication origins that border the so-identified replication domains are replicated earlier than their surroundings whereas the central regions replicate late in the $S$ phase. We discuss the implications of this first experimental confirmation of these replication origin predictions that are likely to be early replicating and active in most tissues.

Figure 1

(a) Skew profile $S$ of a 4.3 Mbp repeat-masked fragment of human chromosome 6; each point corresponds to a 1 kbp window: red, sense ($+$) genes; blue, antisense ($-$) genes; black, intergenic regions (the color was defined by majority rule); the estimated skew profile [Eq. (1)] is shown in green; vertical lines correspond to the locations of 5 putative oris that delimit 4 adjacent domains identified by the wavelet-based methodology. (b) Transcription-associated skew $S_T$ obtained by subtracting the estimated replication-associated profile [green lines in (c)] from the original $S$ profile in (a); the estimated transcription steplike profile [second term on the right-hand side (rhs) of Eq. (1)] is shown in green. (c) Replication-associated skew $S_R$ obtained by subtracting the estimated transcription steplike profile [green lines in (b)] from the original $S$ profile in (a); the estimated replication serrated profile [first term in the rhs of Eq. (1)] is shown in green; the light-blue dots correspond to high-resolution $t_r$ data.

Figure 2

(a) Histogram $N(t_r)$ of the replication timing ratio $t_r$ at the 83 putative oris. (b) Histogram $N(\Delta t_r)$ of the difference $\Delta t_r$ in replication timing ratio between the borders and the center of the 38 replication domains of length $L\ge 1\mathrm{Mbp}$. (c) Percentage of border-center pairs such that $\Delta t_r>0$ (▴) and mean value $⟨\Delta t_r⟩$ (●) versus the domain length $L$; domains have been ordered by length and binned into 6 groups of border-center pairs.

Figure 3

(a) Average replication timing ratio ($\pm \mathrm{SEM}$) determined around the 83 putative oris (●), 20 oris with well defined local maxima (○) and 10 late replicating oris (▵). $\Delta x$ is the native distance to the origins in Mbp units. (b) Histogram of the Pearson’s correlation coefficient values between $t_r$ and the absolute value of $S_R$ over the 38 predicted domains of length $L\ge 1\mathrm{Mbp}$. The dotted line corresponds to the expected histogram computed with the correlation coefficients between $t_r$ and $|S|$ profiles over independent windows randomly positioned along chromosome 6 and with the same length distribution as the 38 detected domains.

Figure 4

Correlation between the replication-associated compositional asymmetry $S_R$ and the replication timing ratio $t_r$ along 5 replication domains bordered by rather early-replicating putative oris. (a)–(e) Phase portrait representation of $t_r$ vs $S_R$; the arrows points in the 5’-3’ direction. (a’)–(e’) Corresponding $S_R$ (black points) and $t_r$ (blue dots) profiles.