The manuscript entitled "Application of multiplet structure deconvolution to extract scalar

coupling constants from 1D NMR spectra" presented by D. Jeannerat and C. Cobas proposes 

an efficient automated solution of the problem of coupling pattern description in 1D ¹H NMR

and extensions to moderately strong couplings and to couplings with nuclei of S > 1/2.

The manuscript was well prepared and is well written.

The doublet to singlet reduction process that constitutes the cornerstone of this manuscript

is based on Figure 2 and some aspects of it are unclear to me.

Convoluting a doublet with something that leaves a singlet at the end is unexpected.

You will find attached a revised Figure 2 (left panel). My red "ovals" show Dirac functions

that are not expected to be there, unless M(J*) is not made of 10 Dirac functions but of an

"infinite" number of them, something that cannot be guessed from Figure 2 (left).

What is reported in the right panel is not equivalent to what is reported in the left panel,

unless I misunderstood the process you reported.

Scanning through the data points and subtracting the current amplitude and the amplitude of the data point J* Hz ahead

in the original spectrum is a way to calculate a convolution product by an anti-phase doublet of Dirac functions.

The convolution of an in-phase doublet(J) with an anti-phase doublet(J) provides an anti-phase

doublet(2J) and is not made a single positive peak, unless some removal of the negative part of the

resulting spectrum is carried out. The present description of the "scrapper" procedure seems very close to

the one of J-doubling, reported by R. Freeman and coworkers.

The only request of the reviewer to the authors is to make Figure 2 more precise and to 

describe the algorithm they use in a clearer and more detailed manner, at least when

it deals only with weak couplings and S= 1/2 nuclei.

A: author statements, C: my comments

A: On page 1, section 30 the author writes: “The sad observation is that the impact of these developments is extremely limited because the broad community of chemists almost exclusively rely on basic 1D 1 H spectra to extract coupling constants.”

C: Indeed, in most cases chemists measure the couplings for reporting NMR spectra of their compounds to provide a “fingerprint” type information where there is no need for particularly accurate J-coupling information, and we can’t blame them for not spending time on learning how to obtain such information with great precision and accuracy. Of course, if more accurate information is made available at their fingertips they will welcome the new tools.

A: On page 2, section 65 the author writes: “In the field of NMR, methods improving the spectral lineshape (relaxation and field inhomogeneity) were developed in the group of Gareth A. Morris under the general 70 framework known as Reference Deconvolution, (Morris et al., 1997;Morris, 2002) but the difficulty to automatize them (i.e. automate the process) makes deconvolution disappointingly disregarded. “

C: I would argue that the proposed technique is likely meet a similar fate as the reference deconvolution. Of course, this does not mean that such techniques should not be developed because clearly there are situations where they make a difference. However, there are always limits of applicability (e.g. SNR) and a price to pay.

A: The author continues: “We should not (i.e. shall not) discuss, here, the deconvolution of spectral lineshape but concentrate on the deconvolution of the effect of coupling interactions called “multiplet-structure deconvolution”.”.

C: Perhaps it is worth mentioning that in many situations combining both techniques may, in fact, improve the results (see e.g. section 150).

A: Second-order effects.

C: The author only considers the symmetry (“roof”) effects. However, there are other second-order effects, for instance the multiplet components may also have unequal linewidth due to various relaxation effects (quadrupolar, CSA etc.). How will this affect the analysis?

A: On page 9, section 295 the author writes: “Finally, a refinement of all parameters: signal intensity, chemical shift, coupling constants, but also roof effect, signal phase, baseline level, and lineshape parameters could be used to further improve the quality of the validation and take into account context-dependent information such as the degeneracy of the coupling constants.”

C: While discussing refinement and validation options the author never considered simulations (e.g. for limited/localized spin systems), which should provide the ultimate means of verification.

A: On page 10, section 335 the author writes: “Combined with other efforts aiming at automatizing (i.e. automating) the analysis of NMR spectra, these methods can significantly increase the quantity and quality of NMR data available to the community. But this will only occur if researchers make their NMR spectra and extracted data available as supplementary data deposited on public databases instead of providing only crude images of spectra and other non-computer readable information.”

C: Surely the chemists will argue that this will occur when the spectra analysis becomes a pushbutton operation. Likewise, the readers of this article will surely agree that there is still a long way to go before this is achieved by analysing only 1D proton spectra. Clearly, some of the problems, mentioned in this article, such as estimating the position of the coupling partners, can be simplified by involving some very basic 2D data. Such a possibility should be at least mentioned.

Generally, the manuscript is reasonably well written, the illustrations are of good quality, the references are appropriate. There are a few typos and phrasing issues, as suggested in the attachment. These, however, only require some minor corrections. Therefore, I am pleased to recommend the manuscript for publication subject to a minor revision.