Synthetic chemistry support

[thorlucas]

Hello,

This is an incredible project. I am amazed at how good this project is compared to other FOSS ELNs.

However, I am a synthetic chemist. This project's defaults seem to be more friendly to bio/physics/analytica/echeml type people. The defaults don't make very much sense at all in my field.

I would love to see this project grow to support the needs of chemists and or other specific fields. Wouldn't it be great if upon setup you could select which field you are in for different sensible defaults? Or perhaps different releases featuring different default templates?

For my field of chemistry, we need:

• Chemical rendering
• Reactions
• Stoichiometry

For an example of how these things should work, look at Signals ELN. It is a perfect example of an ELN for chemists.

Chemical Rendering

This could be accomplished with the support of obsidian-chem. Perhaps when this support is enabled in the settings, the templates could automatically insert chemical drawings for defined chemicals and reactions.

Reactions

Similar to the current "Processes" but it should include reactants, products, solvents, and conditions. It should contain a reaction diagram (again using obsidian chem) and it should have a space for typing the procedure and notes.

Stoichiometry

This is absolutely essential for any chemist, but its a very difficult thing to implement. The way Signals notebook does it is perfect, but it's quite complex. Perhaps this could also be done by a helper plugin and enabled optionally to not add bloat to this project.

---

I have some limited experience with developing personal Obsidian plugins. With some pointers on how we could architect something like this, I could work on tackling this project if there is any interest.

[thorlucas]

I have implemented chemical rendering pretty easily by modifying new_chemical.js to ask for a SMILES string, then adding:

\`\`\`smiles
=this.chemical.smiles
\`\`\`

This requires obsidian_chem.

The next step is to leverage obsidian chem to automatically fetch molar mass and chemical formula by leveraging its RDKit integration.

[fcskit]

Hi,

thanks for the constructive feedback. I moved your post from the “Problems” section to the “Discussions/Ideas” section because, in my opinion, it is not really a bug, but rather a discussion about the further development of the project.

It would definitely be great to expand the ELN to support more scientific domains.

To be honest, when I started the project, I just wanted to create a few simple templates for my students so that they could better document their research. I am still surprised myself at how far this idea has come, and it is a great pleasure to see that other scientists also find the project useful.

I still have many ideas on how to develop the project further. I don't know if you saw that I just released a new version that replaces the property view with a nested properties editor. It's still a dataviewjs view for backwards compatibility, but as such it's a bit slow. So I plan to port it to an Obsidian plugin.

Also, the core functions of the ELN should probably be best ported to an Obsidian plugin. The current version of the templates, which are now mainly JS files and only use the Templater plugin as a trigger, have already been written with this in mind. When porting to an Obsidian plugin, I would also like to add custom modal dialogs. Ideally, these modal dialogs can be designed generically so that different example types and their metadata parameters can be defined by the user in the plugin settings.

My main problem at the moment is that I can only spend a limited amount of time on the project, since I am a scientist myself and developing the ELN is not one of my core tasks. Fortunately, I will be getting a little support from a student assistant starting next month, but to ensure sustainable development of the project, I will probably have to see how I can get institutional funding for further development.

That being said, I am grateful for any help.

[thorlucas]

You're right it was a mistake to put it under "problems", I didn't realize I did that.

Porting core functionalities to a plugin is absolutely the move. This would also make it easier for people to integrate it into existing setups if they don't want the whole template. And it would really clean things up.

Perhaps the plugin could install template files optionally on setup? There could be several "domains" that would have slightly different templates which are useful to different fields. These templates could be auto-generated based on a domain config file, and the user could select an existing one or configure their own template.

I also don't have very much time to play with this as I am in the lab 10 hours a day 😅 but it is a good side project.

I am interested in this project because I would like to keep my personal pet chemistry projects in a separate "forever" ELN rather than my institutional Signals notebook which will vanish when I move to a different institution. I also love the flexibility of Obsidian, so getting something like this working for me would be really awesome.

[thorlucas]

By leveraging obsidian_chem plugin I have used RDKit to auto-calculate the molar mass based on the SMILES string. RDKit also offers a lot of substructure search features which would make fantastic additions to a chemistry ELN.

However, the RDKit js build does not have the ability to create a chemical formula, unfortunately. It is possible to build a version that does do that, but it would be complicated as we would have to maintain our own RDKit JS fork. Bad idea.

I am also considering the possibility of a side-plugin which users can input a personal RSC / ChemSpider API key (1000 free requests per month!) to autofill all this chemical information.

[fcskit]

Thanks for the hint with RDKit but messing around or integrating RDKit into an ELN plugin would probably be Step 4 or higher on the development roadmap. But I agree some of the features would definitely make a nice add-on for the ELN.

For calculating the molar mass it would probably be easier to ask the user to enter the chemical formula (sum or lewis). Then it shouldn't be too difficult to write some code to parse the formula and calculate the molar mass using a lookup dictionary for the element masses.

I don't know also if a SMILES string is the best starting point for this to most users. To me (I'm a materials scientist) this looks rather cryptic. For standard chemicals it would probably the best to obtain the SMILES string from a database, for instance based on the CAS number.

I wrote a quick javascript function to do this

async function getSmiles(casNumber) {
    const url = `https://cactus.nci.nih.gov/chemical/structure/${casNumber}/smiles`;
    try {
        const response = await fetch(url);
        if (!response.ok) {
            throw new Error(`HTTP error! status: ${response.status}`);
        }
        const smiles = await response.text();
        return smiles;
    } catch (error) {
        console.error('Error fetching SMILES string:', error);
        return null;
    }
}
getSmiles('119-61-9').then(smiles => console.log(smiles));

But I couldn't get it to work with the obsidian-chem plugin, since the ```similes ``` environment apparently has only very limited javascript support. Do you know if the smiles renderer can also be called from within a dataviewjs code block?

[thorlucas]

integrating RDKit into an ELN plugin would probably be Step 4 or higher on the development roadmap

I see where you're coming from, integration with cheminformatics toolkits and other plugins etc is complex and best done when the architecture is all sorted out (especially if you will be refactoring/porting things to a plugin). There are definitely more important things to work on.

However, chemical rendering is a pretty high requirement for a lot of chemists, and that requires such a toolkit to be used somewhere in the dependency chain anyway.

Unfortunately, obsidian_chem is not really made to expose RDKit (its just a happy accident it gets added to window), so its not a reliable provider for that (RDKit will not be exposed if a different renderer is selected, for example). So that's pretty hacky for now.

For calculating the molar mass it would probably be easier to ask the user to enter the chemical formula (sum or lewis). […] I don't know also if a SMILES string is the best starting point for this to most users. […] For standard chemicals it would probably the best to obtain the SMILES string from a database, for instance based on the CAS number.

The reason I went to SMILES is because it is the best/fastest way currently to copy a structure from ChemDraw and have it rendered as a drawing in the ELN. Since we have the SMILES already, it makes sense to pull the rest of the information from that.

For reference, my workflow looks like this: I come up with an idea for a molecule I want to make, literature research, draw the reaction up on ChemDraw, and then use an ELN (Signals in my case) to calculate the stoichiometry I will need. Then I take notes in the ELN as the reaction progresses. Finally when I need to write the SI, the ELN already has all the data I need all neat and tidy.

If you are drawing a reaction on ChemDraw, grabbing the CAS numbers is actually a huge hassle. Whereas you can simply draw your chemical and immediately copy the SMILES string with a single keyboard shortcut.

But yes, probably only chemists working with small molecules would really benefit from this. Most other fields of science or even other fields of chemistry would not care much for this.

But I couldn't get it to work with the obsidian-chem plugin, since the similes environment apparently has only very limited javascript support. Do you know if the smiles renderer can also be called from within a dataviewjs code block?

Yes, I was struggling with this too. The way the obsidian_chem plugin is structured does not expose any API.

[fcskit]

I think your workflow can be implemented with small adaptions to the current ELN.

First of all, I think there is a slight conceptional misunderstanding on how you tried to use the ELN. The Chemical template is not ment to be used for new compounds that you plan to or have synthesized. The chemical template is ment to make records of the base chemicals you buy and use as starting materials for your syntheses. A synthesis itself is a process, so it should use the process template. The way I designed the ELN, a process does not describe a specific synthesis but rather is a template. So, let's say you want to carry out a Dies-Alder-Cycloaddition reaction or any other type of chemical reaction. You create a process template for that reaction, with all the devices you need for that reaction (be sure to add them first using the device template) and additional parameters. For your actual synthesis you create a new sample and choose the type compound. During the process of creating a new sample you will be asked to add your educts, which you can choose from your chemicals and samples (in case you want to make a modification of an existing compound), and the process(es). Here you can select your process template for the Dies-Alder reaction or any other process. You can even make multiple selections if your synthesis consists of several synthesis steps. Once your sample (compound) note is created you can fill in the process (synthesis) parameters (temperature, gas atmosphere, etc.) as defined by your process.

What's missing so far is an assignment of educts to the different process steps in case you have a multi-step synthesis procedure. The crucial part to implement this, is how to best represent this in the metadata of the note.

The calculation of molar masses and synthesis weights that you are missing from the Signals ELN could probably be implemented similar to the excel sheet I have created years ago for that purpose. I shared it this morning on my gitHub profile xlsx-compound-calculator. One could probably port this to obsidian either using a dataviewjs view or an obsidian plugin. Using the code from my npe properties editor one could also automatically update the frontmatter accordingly.

[fcskit]

In general, I don't think it's a good idea to base calculations like molar mass, etc. on a SMILES string. Maybe organic chemists get used to reading SMILES strings, but I doubt so. I find it far from easily readable by humans. But it cannot easily be parsed either and implementing a library like RDKit just for parsing a SMILES string to get a molar mass seems to me highly overblown.

Therefore, I think one should always add something like a condensed formula (constitution formula in German) to the metadata, e.g. (CH3)2CHOH. This is easily readable, can be converted into a sum formula rather easily to calculate the molar mass.