This is a computational biology work aimed at modeling the potential mechanisms that govern the formation of retinal cell type mosaics. In brief, the authors define three mechanisms , CF (cell fate), CD (cell death) and CM (cell migration) and determine the regularities of cellular mosaics resulting from implementations and parameter adjustments of the three mechanisms. In general the work is thought-provoking, but, given a certain lack of clarity of presentation, and many oversimplifications of the actual biological processes and definitions, its relevance to the actual biological process are somewhat unclear to me.

I will make several comments on the methodology and results, and end by questioning nearly every conclusion of the paper:

Line 43-47: "RGC type classification criteria":
4 = Regular spacing is indeed a criterion, but it seems to me it is more relevant to evaluate the density of arbors / coverage factor rather than the regularity of cell body spacing. An inspection of the eyewire museum shows that cell bodies of morphologically identified cell types are by no means regular in spacing, but rather exhibit strong excentricities from the centroid of the dendritic arbor, while the dendritic arbors are the true measure of tiling and coverage of the retina. The picture of regularly spaced SAC cell bodies, in its nearly christaline beauty is also misleading, as SACs exhibit an unique and unusual regular dendritic arbor with cell bodies almost always perfectly centered on the centroid.
Additionally, central projections which link into the function of RGCs are major classification criteria of cell type (see for instance Chen, Badea Hattar 2011, where ipRGCs with identical properties within the retina seprate into two subpopulations, projecting to two different retinorecipient nuclei and subserve two dramatically different functions, pupil constriction and circadian rythms , hence by all means suggesting two distinct cell types).

Line 53-54: "Interestingly, not much is known yet about the impact of RGC apoptosis on the
maturation of retinal circuitry and visual pathways" - I would recommend the authors to read about the effects of cell death on RGC type spacing in: "Apoptosis regulates ipRGC spacing necessary for rods and cones to drive circadian photoentrainment" Chen et al 2013 Neuron. Figure 1 of that paper is particularly relevant to the theoretical framework of their work.

Line 59-67 : Cell fate determination:

This paragraph is highly problematic to me. The referenced work does not show examples of Cell Type desicions (e.g. ONalpha RGCs vs. betta RGCs) , but at best points to distinctions between cell classes (Amacrine vs. Bipolar vs. RGCS) which are not very relevant for the proposed modeling work. A recent example from the mamalian literature (albeit indirect) where a signaling cascade influences cell fate is Muzyka et al 2021, where Ret signaling is implicated through genetic manipulations to be required for RGC cell type specification. However the best examples are coming from the fly literature.

Line 91: "In all cases, cell-cell interactions seem to be mandatory for tangential migration."
This is in my view a very strong statement that is not supported at least through the cited work.

Line 200: Substance Secretion:
In this paragraph and in general, the paper would benefit from a more clear, (perhaps graphic summary) description of the algorithm.
iWhat does the concentration do? What is the threshold past which a target precursor is prohibited to go down that path ? What is the decision logic of an uncomited precursor to engage in any given cell type path? The "least inhibited path?"
Line 210 : Cell death

It is not clear from biological literature what the balance between "cell death" or "alternative cell fate choice" is. However the authors can and should define more precisely the rules.
Is the inhibitory substance deviating cell fate at one threshold and killing the cell at a different one? Again a graphic representation of the various outcomes for one progenitor should be presented in order to make the work clearer.

Line 213:
how is "naive" cell generation controlled? from a set pool of precursor that gets eventually exhausted? or endless ?

Line 264:

T-tests are conducted. Are the underlying compared observations normally distributed? How many samples are being compared? Are tests that do not assume normal distribution more appropriate?

Line 298:

Figure 3 and corresponding results sections. A central claim of this study is that the CM mechanism is the most significant contributor to cell regularity (cell body that is). However there is no comparison of CM alone to CF, CD or CF+CD. If one infers from the presented graphs, it results that actually CM contributes only a minimal improvement, after most of regularization has been implemented by CF and CD. In figure 3 E, for instance, the RI goes from under 2 to 3.5 at the end of CD, only to rise to 4 after the CM component. A direct comparisons of the mechanisms should be provided.

Line 394: Figure 7

It is surprising to me that cell density of SACs is not changing between P4 and P10, while the diameter of the retina is clearly increasing. Does this mean that there are new SACs born into the mosaic during that time?

Lines 384 - 388:

I have a hard time following the argument about treating ON and OFF SACs as one population.
The inference that their spacing is not changing after P4 ("shortly after GCL and INL separation") is not very strong, as the separation of the GCL and INL is already in place at E17 (DAPI staining), and the IPL is in place at P0. IN fact SACs of the INL and GCL can be distinguished with markers such as Isl1 as early as E17.
Moreover, the exclusion argument is not clear to me: Figure 6 shows to nicely overlapping ON and OFF SAC mosaics, where most cell bodies appear nicely spaced and close to each other across layers, almost begging teh question of whether tehy are not born pairwise from precursors. perhaps a clear quantitation of the nearest OFF vs. ON neighbor for each ON cell and viceversa would be informative?

Now to the conclusions of the paper:

1) "We report that the cell migration mechanism yields the most regular mosaics." Clearly not proven to me. Perhaps, the CM mosaic contributes significantly to the regularity of the mosaics ? See comment above. How would the RI look like if only CM mechanism would be implemented ?

2) "We report that homotypic interactions between the GCL and INL populations during mosaics creation are required to reproduce the observed SAC mosaics’ characteristics"

This would be a strong statement even for a wet-lab paper. "We suggest/predict/infer/propose" is far more in line with the presented evidence.

I think the authors have done the best possible job within the framework of their system.