Scientists pride themselves on clear, logical and concise communication. So naturally, the process for publishing our research involves an absurd number of formalities, like coming up with 700 slightly different ways to ‘thank the reviewer for their insightful comment’. Nevertheless, I’m told this is all a necessary part of spreading your beautiful researcher butterfly wings\u2014and frankly, I’m enough years into my DPhil to stop questioning every quirk of academia. However, the current protocol for new researchers wanting to learn the moves to this bizarre dance seems to be begging postdocs\/ old timers for examples of cover letters, marked-up manuscripts, and reviewer responses. To attempt to save everyone some time, I thought I’d provide some guidance and templates here.<\/p>\n\n\n\n\n\n\n\n
First, a few disclaimers:<\/p>\n\n\n\n
Now I’m legally absolved, let’s get into the horror show<\/s> fun stuff!<\/p>\n\n\n\n
Step 1: choosing a journal<\/strong><\/p>\n\n\n\n You’ve finally reached the point where you and your supervisor(s) have agreed the project is ready to be written up. Congratulations! Savour this moment, because it\u2019s the start of a long, winding road to publication.<\/p>\n\n\n\n Before you even begin writing, you have an important first task: deciding which journal to target. It\u2019s tempting to write first and choose later, but having to reformat your entire manuscript and all its figures is a uniquely painful experience. Having a primary journal in mind from day one helps you write for the right audience and format.<\/p>\n\n\n\n So, how do you choose? Start by asking yourself a few key questions:<\/p>\n\n\n\n After considering these, you’ll likely narrow it down to a handful of options. For those of us in OPIG, a few familiar names that come to mind are Journal of Chemical Information and Modeling (JCIM), Chemical Science, Digital Discovery, PNAS, Bioinformatics, and the Nature Communications and Frontiers families.<\/p>\n\n\n\n Step 1.5: submitting to arXiv<\/strong><\/p>\n\n\n\n Wow – you’ve not only written up your work, but also got it through the toughest reviewers of all: your co-authors! With their go-ahead, you can save that 17th draft with great pride and get ready to upload it to a preprint server (probably bioRxiv). I actually think uploading your paper here might be the scariest part of the whole process; after all, it’s the first time complete strangers can read your work and, more frighteningly, have critical thoughts about it if they want (I’ve never actually received any nasty comments, but nightmares gotta nightmare). I\u2019ve included this as step 1.5 because preprinting isn’t required, and sometimes there might be reasons not to do it, but if you do, it’s worth doing the social media-ing from Step 6 at this point too. Just remember to let the group’s social media people know so they can repost\/share.<\/p>\n\n\n\n Step 2: formatting, portals, and choosing reviewers<\/strong><\/p>\n\n\n\n Now, manuscript in hand, it’s time for the administrative slog. This is the point where you format your work to the journal’s exact specifications. Hopefully, because we planned ahead in Step 1, this won’t be a complete nightmare\u2014even so, expect a bit of a faff.<\/p>\n\n\n\n Once it looks the part, you’ll get to wrestle with the submission portal. These websites\u2014usually some version of ScholarOne or Editorial Manager\u2014often feel like they were designed in the early 2000s and never updated. Here\u2019s a pro-tip: way before you’re actually ready to submit for real, do a ‘dry run’: click through as much of the portal as you can without hitting the final button. This way, you’ll know exactly what you need to gather. You’ll uncover all sorts of things you’ll probably want to check with your supervisor, like which publishing licence to choose, whether to suggest a specific editor, whether you need a graphical abstract, or where to find the exact grant funding numbers.<\/p>\n\n\n\n The final, and most strategic, part of this stage is choosing your preferred reviewers. Most journals ask you to suggest 3-5 people who could competently review your paper. Don’t overthink it, but do be sensible. A good place to start is once again your reference list: who are the authors of the key papers you\u2019re building upon? You’ll often have to provide email addresses for your suggestions, and as far as I know the best way to do this is online stalking, and failing that, relying on generic address formatting to deduce them (eg: Prof Tinky Winky at Teletubby University is likely reachable at tinky.winky-at-teletubby.ac.uk)<\/p>\n\n\n\n Step 3: writing a cover letter<\/strong><\/p>\n\n\n\n As well as filling out the portal shenanigans, you’ll also have to prepare a cover letter\/ letter to the editor introducing your paper, highlighting your key findings, and explaining why it is interesting to the journal. An example of one of these is below:<\/p>\n\n\n\n Dear Editors,<\/p>\n\n\n\n Please find enclosed our original manuscript titled “An Investigation into the Surprising Fluid Dynamics of Tubby Custard.”<\/p>\n\n\n\n In this work, we address a key challenge in cafeteria science: the inconsistent dispensing of Tubby Custard. By systematically analysing the custard’s properties under various conditions, we have identified the key factors that control its viscosity and flow. Our findings provide a framework for ensuring a more reliable and enjoyable custardy experience, minimising unexpected splashing and maximising deliciousness.<\/p>\n\n\n\n We believe that this work would be an excellent fit for your journal and of great interest to your readership.<\/p>\n\n\n\n Thank you in advance for your consideration.<\/p>\n\n\n\n Kind regards,<\/p>\n\n\n\n Tinky Winky, Dipsy, Laa-Laa, and Po<\/p>\n<\/blockquote>\n\n\n\n Whimsical though I’ve (hopefully) made it seem, you do want to make sure you get this right, so absolutely get it checked over by a co-author.<\/p>\n\n\n\n Step 4: pressing the button and w a i t i n g<\/strong><\/p>\n\n\n\n You\u2019ve navigated the portal, uploaded everything but the kitchen sink, and finally hit ‘submit’. Now the waiting begins. The time until the first decision varies enormously. Most journals have a tracker in their portal to show what stage your manuscript is at (e.g., ‘With Editor’, ‘Under Review’), and some will email you with updates. If months go by with no change, it might be worth asking your supervisor whether a polite nudge to the editor is appropriate. Otherwise, try to put it out of your mind and enjoy having the brain space to think about your next project.<\/p>\n\n\n\n Step 5: receiving reviews and responding to reviewers<\/strong><\/p>\n\n\n\n Common outcomes are rejection by editor, rejection by reviewers, major corrections, minor corrections and accepted. Major corrections, though, seems to be the most common of the good outcomes, and it’s worth stressing that even though it sounds scary, this is a good outcome! The reviewers found your work interesting enough to engage with it and give you some hopefully interesting feedback. Now you can get started with making some of these changes, but not a second before you do two things:<\/p>\n\n\n\n 1.Set up your response to reviewers document.<\/strong> This is the very first thing I do; copy and paste all of the comments into a google doc (\/editor of your choice). You’re eventually going to add a little introductory paragraph, and responses to each of the individual reviews, all colour coded. These responses should contain either a defense of the point (we haven’t changed this, here’s why), or an explanation of what you’ve changed. Beneath the explanation, you’ll also be expected to quote the exact part of the manuscript has changed. This is easier to show than explain, so see below an example:<\/p>\n\n\n\n Deaar Professor Noo-noo,<\/p>\n\n\n\n Please find enclosed our updated manuscript titled “An Investigation into the Surprising Fluid Dynamics of Tubby Custard”. We are grateful for the valuable feedback provided by the reviewers, which has helped us to significantly improve the paper. Below we give a point-by-point response, with the reviewers’ comments in blue, our responses in black, and all changes to the manuscript in red.<\/p>\n\n\n\n Kind regards,<\/p>\n\n\n\n Tinky Winky, Dipsy, Laa-Laa, and Po<\/p>\n\n\n\n The Home Dome Institute for Applied Snacking, Teletubbyland<\/em><\/p>\n\n\n\n This is an interesting paper that addresses very timely issues in the use of AI for cafeteria science. It is likely to be of interest to readers of the Journal of Applied Snacking<\/em>. Detailed comments and suggestions follow.<\/p>\n\n\n\n The Introduction focuses on the challenge of avoiding ‘recipe memorisation’, but it is not clear what steps in the Methods section are designed to avoid this problem.<\/p>\n\n\n\n We thank the reviewer for pointing this out, and agree that this could be made clearer in the manuscript. We have added the following in the Methods section of the paper:<\/p>\n\n\n\n As mentioned above, a prominent problem in custard-based machine learning is the memorisation of dataset biases. The effects of such biases can be reduced by limiting ‘recipe leakage’ between the train and test datasets. To this end, we prepared filtered subsets of both the ‘Standard Pink’ and ‘Festive Green’ batches for training and testing, as well as random subsets of the original training sets of the same size as the filtered sets.<\/p>\n\n\n\n I may have missed it, but I didn\u2019t notice how the \u2018Custard-System\u2019 model was constructed. I presume each node is a Teletubby, but are the edges only physical proximity, or do they also represent excitement levels less than e.g. 6 ‘Again-Agains’? This information may be in the Secret Instructions, but it seems important enough that it should be briefly stated in the main text.<\/p>\n\n\n\n We thank the reviewer for bringing this to our attention, and agree that this would be useful information for readers. We have moved the following description of the model construction process from the Secret Instructions to the main body of the paper:<\/p>\n\n\n\n Our model takes four inputs: positions, node embeddings, edge indices, and edge embeddings. If there are n Teletubbies, the positions tensor contains the x, y and z coordinates of each Teletubby. The node embeddings are a tensor of one-hot encoded Teletubby types, with a bit to distinguish between who is holding the bowl and who is not. Edges are generated on-the-fly using a variable excitement cutoff. We use a cutoff of 10 hugs for Teletubby-to-custard-machine edges, and 2 hugs for Teletubby-to-Teletubby edges. In this way, inter-Teletubby edges closely mimic social bonds, with a much more expansive connectivity to describe interactions with the custard dispenser.<\/p>\n\n\n\n Page 8, line 30: \u201cdue to the lack of tastiness data\u2026\u201d I got a little lost here. Surely Teletubbyland has plenty of tastiness data. What is the limitation alluded to here?<\/p>\n\n\n\n We thank the reviewer for highlighting this. We are referring specifically to structured data linking a dollop’s physical form to its yumminess score, and agree that it is unclear in the text. To resolve this, we have added a clarification in the line specified:<\/p>\n\n\n\n Due to the lack of tastiness data for dollops with known physical structures, we choose to pre-train on a ‘dollop classification’ task in order to try and learn some of the physics of deliciousness before finetuning on yumminess data.<\/p>\n\n\n\n A substantial concern is that (if I am interpreting right) the so-called rigorous filtering allows flavour similarities up to 0.8 ‘Tubbyspice’ Index and batch similarities up to 80% ‘Pinkness’ identity. This is still a lot of similarity! I would suggest rerunning selected studies with tighter thresholds.<\/p>\n\n\n\n We thank the reviewer for this important comment and we are in complete agreement with the concern that they raise. However, creating stricter filters on the custard batches is challenging. If we move the ‘Pinkness’ threshold to 30% (a potentially more realistic cutoff), almost half of the dataset is removed, as most Tubby Custard is, in fact, pink. This reduction in training data would result in a very large drop in performance, making it very difficult to assess how much of this drop could be attributed to the reduction of recipe leakage versus the simple lack of data. We have included the following in the paper to better clarify our aim:<\/p>\n\n\n\n The filtered datasets were constructed by removing custards if they met either of the following criteria: 1. ‘Tubbyspice’ Index between the custard and any of the 285 test set custards greater than 0.8. 2. ‘Pinkness’ identity between the custard batch and any of the 285 test set batches greater than 0.8…<\/p>\n\n\n\n We have also amended our discussion as follows:<\/p>\n\n\n\n Furthermore, although our dataset filtering has gone some way to reducing the bias that our model learns, it is unlikely that we have removed all dependence on it…<\/p>\n\n\n\n We have also removed all appearances of the word “rigorous” when referring to this filtering, as we agree with the reviewer that it could be misleading to readers.<\/p>\n<\/blockquote>\n<\/div><\/div>\n<\/div>\n\n\n\n\n
\n
\n