Paperonium

← Back to the plain-language case

Abstract says PZ DNA has greater twist; the paper's tables and Results show it 12° lower

Nucleic Acids Res (2017) · doi:10.1093/nar/gkx144 · PMC5397145

Summary

The abstract says that in 50 μs MD simulations the duplex with six consecutive P:Z pairs shows "greater average values of stagger, slide, rise, twist and h-rise" than a G:C control. The paper's own data show the opposite for twist. In Table 3, mean twist is 21.6 ± 15.2° for PZ/PZ and 33.6 ± 4.3° for GC/GC, so PZ is 12.0° lower. H-twist is 22.1 versus 34.3°, or 12.2° lower. The Results text agrees with the tables ("about 10° smaller"), and so do the crystal values in Table 2 and every step of the substituted region in the supplement. The abstract is the only place that gives this direction. It is also the part most readers and indexing services pick up, so a reader could come away believing that PZ-rich DNA is more twisted than natural DNA, when the paper's results show less twist and a partly A-like helix.

What the paper reports

Abstract:

In these simulations, the PZ-containing DNA duplex exhibits a significantly wider major groove and greater average values of stagger, slide, rise, twist and h-rise than observed for a ‘control’ oligonucleotide in which P:Z nucleobase pairs are replaced by G:C.

Results and Discussion, "Time convergence and statistics of structural parameters in the MD simulations", paragraph 2:

On the other hand, the mean values for twist and h-twist are about 10° smaller on average for PZ/PZ than GC/GC pairs, with a greater variance (Table 3).

The same paragraph, on stagger and slide:

with the exception of stagger, which is twice as large for the PZ than the GC helix

the average ‘slide’ value is three times larger for the oligonucleotide containing six consecutive P:Z nucleobase pairs compared with that of the control oligonucleotide

The same section, paragraph 3:

In contrast to GC, distributions for PZ are bimodal for slide, twist and h-twist. The values associated with the larger peak in each case are within the range of values reported for A-form.

Table 3, "Helix parameters calculated from the MD trajectories of the PZ and GC-containing DNA duplexes" (excerpt, mean ± SD):

PZ/PZ | … | Twist (°) 21.6 ± 15.2
GC/GC | … | Twist (°) 33.6 ± 4.3
PZ/PZ | … | H-Twist (°) 22.1 ± 16.4
GC/GC | … | H-Twist (°) 34.3 ± 4.2

Table 3, footnotes a and b:

Precision bounds are based on a 95% confidence interval with respect to the standard error of the mean (<0.08 Å and 1° for lengths and angles, respectively).

We have arbitrarily chosen to list only GC/GC and PZ/PZ dinucleotide steps; values for GG/CC and PP/ZZ are given in Supplementary Tables S5–S8.

Table 2, "Comparison of experimental and computational PZ and GC Helix parameters" (central step):

H-Twist (°) | PZ Exp. 29.4 | PZ Comp. 22.1 | GC Exp. 45.0 | GC Comp. 34.3

What the paper's own numbers show

Method. I took the eight PZ/PZ and GC/GC parameters in Table 3 that bear on the abstract sentence and computed PZ minus GC, both as signed values and as magnitudes, so that a magnitude reading of "greater" is tested as well. I then checked the same comparison in Table 2 (MD and crystal), and at each of the seven dinucleotide steps of the substituted region in Supplementary Tables S6 (twist) and S7 (h-twist), matching each PZ step to the GC step at the same position. All arithmetic is in recompute.py.txt.

Quantity Location Reported Recomputed from the paper's own numbers
Direction of twist, PZ vs GC Abstract "greater average values of … twist" the tables show the opposite (rows below)
Twist, PZ/PZ vs GC/GC (MD) Table 3 21.6 ± 15.2° vs 33.6 ± 4.3° PZ − GC = −12.0° (lower, signed or by magnitude)
H-twist, PZ/PZ vs GC/GC (MD) Table 3; Table 2 "Comp." 22.1 ± 16.4° vs 34.3 ± 4.2° −12.2° (lower)
H-twist, central step (crystal) Table 2 "Exp." 29.4° vs 45.0° −15.6° (lower)
Size of the twist difference in the text Results, paragraph 2 "about 10° smaller" 12.0° (twist) and 12.2° (h-twist); same direction as the tables
Twist, all seven steps of the substituted region Tables S6 e.g. PP/ZZ 22.6 vs GG/CC 30.0; ZA/TP 28.8 vs CA/TG 30.3 −1.5, −3.8, −7.4, −12.0, −6.6, −3.8, −1.5°; PZ lower at 7 of 7 steps; mean −5.2°
H-twist, all seven steps Table S7 e.g. PP/ZZ 24.6 vs GG/CC 31.1 −2.5, −3.4, −6.5, −12.2, −5.7, −3.5, −2.4°; PZ lower at 7 of 7; mean −5.2°
Twist difference against the stated precision Table 3 footnote a (1° for angles) — 12 times the precision bound; 1.07 pooled SDs (pooled SD 11.17°)
Twist spread Table 3 SD 15.2° vs 4.3° variance 12.5 times larger for PZ
Stagger Table 3; Abstract; Results "twice as large" −0.19 vs −0.08 Å magnitude larger (ratio 2.38); signed value lower (−0.11 Å)
Slide Table 3; Abstract; Results "three times larger" −2.86 vs −0.78 Å magnitude larger (ratio 3.67); signed value lower (−2.08 Å)
Rise Table 3; Abstract 3.29 vs 3.19 Å +0.10 Å (greater)
H-rise Table 3; Abstract 3.11 vs 3.08 Å +0.03 Å (greater)
Major groove width Table 3; Abstract 26.8 vs 21.0 Å +5.8 Å (wider)

Of the five parameters the abstract calls "greater", rise and h-rise are greater as signed values, and stagger and slide are greater in magnitude, since both are more negative for PZ. Twist is the only one that is lower under either reading. Twist values are all positive, so the sign question does not arise for it.

Figure

The abstract says PZ twist is greater; the Results text and Tables 2 and 3 give lower values for PZ

Figure 1. The two places where the paper states the direction of the PZ versus GC twist difference. Left (orange): the abstract, which lists twist among the parameters with "greater average values" for the PZ duplex. Right: the Results sentence and the Table 3 and Table 2 values for the same comparison, as printed. The blue line at the bottom of the right panel gives the differences recomputed from those table values: twist −12.0° and h-twist −12.2° (PZ minus GC). The text excerpts are quoted verbatim; the table rows are shortened to the twist and h-twist columns.

At every step of the substituted region, PZ twist and h-twist are below the matching GC values

Figure 2. Differences between PZ and GC twist (filled circles, solid line, Supplementary Table S6) and h-twist (open squares, dashed line, Supplementary Table S7), recomputed for each dinucleotide step of the substituted region, with the PZ step matched to the GC step at the same position. The horizontal line marks no difference. The shaded band above it is where the values would lie if the abstract's "greater twist" were right. All fourteen differences are negative, from −1.5° at the flanking TP/ZA and ZA/TP steps to −12.0° (twist) and −12.2° (h-twist) at the central PZ/PZ step reported in Table 3. So no choice of step or average within the paper's data gives a greater twist for PZ.

Why it matters

Explanations considered

Status

Suggested correction

In the abstract, replace

greater average values of stagger, slide, rise, twist and h-rise than observed for a ‘control’ oligonucleotide

with

larger magnitudes of stagger and slide, greater average rise and h-rise, and lower average twist and h-twist than observed for a ‘control’ oligonucleotide

This keeps the abstract consistent with Table 3 and with the Results text. If the authors prefer a shorter change, removing "twist" from the list of "greater" parameters and adding "and smaller twist" would correct the direction. The Results text could also give the twist difference as about 12° instead of "about 10°", to match Table 3 (12.0° for twist, 12.2° for h-twist).