Canonical SMILES generated _within_ the same software package for the same molecule will always be identical. Open Babel (open source software) has one implementation ( ). Nice writeup and a good illustration of an important limitation of SMILES.Īs suggested in the comments, there is something called “Canonical SMILES”. I wonder if a simpler approach to mechanism would be to present the question with all bonds and atoms numbered/alphabetised, with the student asked to enter, as text, the “from” and “to” identifiers for each curly arrow they propose. If you wanted a student to draw a molecule *and* give a mechanism on that molecule, things would be complicated by a need to canonicalise the atom IDs before checking curly arrows. I think you would be able to extract all molecules and “shapes” (arrows, etc) from the drawing applet and then check the “connectivity” described by the curly arrows. For mechanism etc., there may be more of a challenge. Simple structure sketching and answer checking should be quite doable with this. If you wanted to do some hacking on Moodle, there might be a way to incorporate ChemDoodle Web Components, which is freely licensed ( ). The InChI string is also quite a bit longer than either SMILES or InChI Key. This InChI representation is human readable, but there’s a bit of a learning curve. The InChI itself (not the key) should be canonical. There are some “Canonical SMILES” algorithms which should produce consistent representations for the same structure, though ChemDraw apparently doesn’t support them. This is a common-ish issue with many SMILES implementations, as I understand it. If anyone has any suggestions, let me know! It needs to be something unambiguous that the students can use and the non-chemistry quiz grading software can interpret the answer.
I still haven’t figured out a convenient way to code arrow-pushing mechanisms or multi-step synthesis answers using this method. It’s not human readable, but at least the same structure seems to give the same InChI Key regardless of how the structure is drawn. So now, I’m going to use InChI Key for my grading system. I understand that all of these are valid SMILES, but I feel that one structure should give the same SMILES from ChemDraw, regardless of how the molecule happened to be drawn. I’ve let the ChemDraw people know, and they said they’d pass my comments along to developers. Don’t believe me? Watch this, and try it yourself: Start in the middle, and you get a third SMILES still. If you start from the left side, you get one SMILES, from the right you get a different SMILES. jpg structures, all of the structures appeared to be correct! What is going on?Īfter some testing of my own, I came to a disappointing conclusion: ChemDraw gives different SMILES for the same structure depending on the order in which you draw the bonds. All they had to do was follow along bond for bond with my tutorial video and get the correct answer. The first thing I noticed when looking at the responses is that most of the students got most of the tutorial questions incorrect! I didn’t understand why. jpg version of the structure so I could check it visually myself if there were any discrepancies. For testing purposes, I also had them upload a. I coded the correct SMILES into the grader – as well as some of the common errors for feedback purposes – and released it to my students. So I wrote up a pilot quiz with tutorial videos. Perfect! I like the human readable aspect, as students can glance through their SMILES to check for some potential errors. SMILES is a condensed version of the structure, almost human readable, and ChemDraw can copy a structure directly in SMILES format.
And our ChemDraw license does not include the structure-to-name feature. But many of the structures have IUPAC names which can be rather large and/or outside of the scope of IUPAC nomenclature my students learn. So in some respects, I’m admittedly reinventing the wheel here.Īt first I thought the LMS would be able to grade my students’ work if they type in the IUPAC name for each structure. I know some publishers have platforms for just this, but our university doesn’t subscribe to the service. NOT what my exam questions will look like…