Overall, I had some critical comments earlier, but I'm going to try to use Molydyn but will need to implement all the LAMMPS stuff which will require figuring out how to set up LAMMPS on Windows. Molydyn figures out where I'm going to get the CPU power from.
I see the potential for a machine I can stuff coins into and it spits out realistic models in a .xyz or .cdx file for structures and a .csv of properties that I can use to get a picture of why a structure has specific properties rather than just handwaving at sterics or electronics or 'VDW forces go brrr'.
I can incorporate that 'less-than-a-blind-guess insight' into tech sales and planning of new product dev. The service as is only gets you the middle part of the process and it could be a tough sell (low value add) to market this service to hardcore comp chemists who are probably have a career/thesis invested in doing it their own specific way.
There is a big overlooked thing here about Molydyn having a foothold to take LAMMPS and MD to the masses where you would have a larger customer base to support doing more cool things. DFT models - for all its criticism - made a splash in quantum models because it was less scary and easier to pick up, it democratized the field though it may have traded some accuracy and 'commoditized' the process a bit (everybody doing wonky DFT calculations in their synthesis paper). But at least it got more people taking about the field than if we were still running pure ab initio calculations.
Also, the Purchase Credits tab is very hard to understand. What's going on here when I put in 1, 10, 100, etc.? One credit = how many dollars. And then there's apparently some free credits for buying specific numbers of credits?
Apr 11, 2023·edited Apr 11, 2023Liked by Tony Maiorana
Computational polymer scientist here.
Really cool to see Molydyn's work. However, so far it just seems to set up simulations, which other commercial software already do (even if Molydyn might be a bit newer/more modern). But I think there is still a mismatch between what Molydyn (or any of these commercial software) are trying to sell (simulations for non-experts) vs software that makes experts' life easier.
Another question that is often elided is that running simulations is different from running *good* simulations. Simulations are just computer experiments. For example, many "casual" simulation papers in the literature using currently existing commercial software are often ridiculously short. At some point, it's not so much that the computational resources themselves are expensive, but that the *time* to do quality simulations that give meaningful results is expensive. There is also a lot of polymer theory that goes into properly extrapolating simulations into real-life polymer systems.
If someone here does have experience getting good use out of say, Schrodinger, Biovia, Synopsys, or Siemens' molecular dynamics software, I'd love to hear about how simulation has helped your work. I'm especially interested if you're an experimentalist who has benefited from simulations partnership. My experience is that this usually requires really close partnership between a theorist and experimentalist to find the intersection between what is possible via "in vivo" and "in silico" experiments.
Apr 12, 2023·edited Apr 12, 2023Liked by Tony Maiorana
Hey Tony, thanks for all of your great content.
Something that occurred to me: you've given us really great overviews into broad level business decisions and considerations within chemical companies. I'm curious what chemists specifically spend most of their time on?
E.g. I imagine that a lot of process design/optimization draws on chemical engineering, and characterization and formulation design handled perhaps by rheologists/physicists. Meanwhile, it seems that chemists are uniquely poised to think of new monomers/polymers (e.g. what I believe most PhDs in chemistry are about). Is this sense correct? Or is designing new molecules too hard of a business proposition, so chemists spend most of their time devising new ways to make the same thing?
well, I don't want to speak for most chemists, but in my experience I only spent 20-30% of my time actually doing experiments and getting to manipulate data or plan things out.
As a polymer chemist I was often also the rheologist (not common, I self-taught under some guidance from a thesis committee member) and the formulator and the applications tester. I was very comfortable with linear rheology, I even did a company-wide training once at Hexion, but anything that was non-linear I'd prefer a real rheologist to figure it out, but those are kind of few and far between at a chemical company.
I think this could be a post of its own, but most chemists right now I think are expected to handle more and more of the jobs of their co-workers. This might mean plant support (they fucked up and need you to figure out how to rework off-spec material), customer support (product isn't working correctly), process improvements (e.g., batch time from 24 hours to 8 hours), and various cost reduction projects. That's in addition to "figure out that next product that will bring in new revenue."
I think most chemists wish they could spend time coming up with new stuff, but there just usually isn't enough time in the day or resources at your disposal to do it. Also, plenty of "by-product" out there that could be gotten for cheap and turned into new polymers (BPA distillation bottoms, vanillin distillation bottoms, asphalt like stuff, etc)
If you have access to them, I only gave it as an example because it was non-obvious, but I think you can also just view organic chemistry as a tool in the toolbox of the chemist too.
Apr 11, 2023·edited Apr 11, 2023Liked by Tony Maiorana
If you're into books and spreadsheets, I love Van Krevelen's 'group contribution' method for predicting properties of polymers but it's not as inspiring as a whole simulation.
The Molydyn UI needs some work. It's a bit too raw to invite a casual user and something close to click/drag interface like Chemdraw would be more approachable to a pure synthesis person looking to get started in simulation. Someone completely new isn't going to know what LAAMPS is or Leonard-Jones potentials, there has to be some hand holding and oversimplification with a JMol kind of input/output for most, and an Advanced Mode for experienced MD users who could enter all the parameters in just a DOS prompt if they wanted.
I should probably watch their tutorial haha. Even Gaussian for DFT or Abalone for MD are easier to hack around in and at least get something wiggling on the screen.
Apr 11, 2023·edited Apr 14, 2023Liked by Tony Maiorana
Thank you for the honest feedback! You're spot on with the need to simplify things - right now we're catering to computational chemists who've already got experience with LAMMPS. The update that's coming soon will be our first step into supporting all chemists. Do get in touch at hello@molydyn.com and I'd be happy to give you a demo! Thanks, Matt
Apr 11, 2023·edited Apr 12, 2023Liked by Tony Maiorana
I don't think 'able to be accessed' (free trial access) and 'accessibility' (user competency in workflow and interpretation) parse out to the same thing in the context of high level physical chemistry.
Shouldn't a practicing computational chemist already have a terminal to a corporate or university server or Amazon Web Service?
Sorry, mistake on my part with the contact email.
Please use hello@molydyn.com if you want to reach out to Matt.
Overall, I had some critical comments earlier, but I'm going to try to use Molydyn but will need to implement all the LAMMPS stuff which will require figuring out how to set up LAMMPS on Windows. Molydyn figures out where I'm going to get the CPU power from.
I see the potential for a machine I can stuff coins into and it spits out realistic models in a .xyz or .cdx file for structures and a .csv of properties that I can use to get a picture of why a structure has specific properties rather than just handwaving at sterics or electronics or 'VDW forces go brrr'.
I can incorporate that 'less-than-a-blind-guess insight' into tech sales and planning of new product dev. The service as is only gets you the middle part of the process and it could be a tough sell (low value add) to market this service to hardcore comp chemists who are probably have a career/thesis invested in doing it their own specific way.
There is a big overlooked thing here about Molydyn having a foothold to take LAMMPS and MD to the masses where you would have a larger customer base to support doing more cool things. DFT models - for all its criticism - made a splash in quantum models because it was less scary and easier to pick up, it democratized the field though it may have traded some accuracy and 'commoditized' the process a bit (everybody doing wonky DFT calculations in their synthesis paper). But at least it got more people taking about the field than if we were still running pure ab initio calculations.
Also, the Purchase Credits tab is very hard to understand. What's going on here when I put in 1, 10, 100, etc.? One credit = how many dollars. And then there's apparently some free credits for buying specific numbers of credits?
Computational polymer scientist here.
Really cool to see Molydyn's work. However, so far it just seems to set up simulations, which other commercial software already do (even if Molydyn might be a bit newer/more modern). But I think there is still a mismatch between what Molydyn (or any of these commercial software) are trying to sell (simulations for non-experts) vs software that makes experts' life easier.
Another question that is often elided is that running simulations is different from running *good* simulations. Simulations are just computer experiments. For example, many "casual" simulation papers in the literature using currently existing commercial software are often ridiculously short. At some point, it's not so much that the computational resources themselves are expensive, but that the *time* to do quality simulations that give meaningful results is expensive. There is also a lot of polymer theory that goes into properly extrapolating simulations into real-life polymer systems.
If someone here does have experience getting good use out of say, Schrodinger, Biovia, Synopsys, or Siemens' molecular dynamics software, I'd love to hear about how simulation has helped your work. I'm especially interested if you're an experimentalist who has benefited from simulations partnership. My experience is that this usually requires really close partnership between a theorist and experimentalist to find the intersection between what is possible via "in vivo" and "in silico" experiments.
Loving this feedback
Hey Tony, thanks for all of your great content.
Something that occurred to me: you've given us really great overviews into broad level business decisions and considerations within chemical companies. I'm curious what chemists specifically spend most of their time on?
E.g. I imagine that a lot of process design/optimization draws on chemical engineering, and characterization and formulation design handled perhaps by rheologists/physicists. Meanwhile, it seems that chemists are uniquely poised to think of new monomers/polymers (e.g. what I believe most PhDs in chemistry are about). Is this sense correct? Or is designing new molecules too hard of a business proposition, so chemists spend most of their time devising new ways to make the same thing?
well, I don't want to speak for most chemists, but in my experience I only spent 20-30% of my time actually doing experiments and getting to manipulate data or plan things out.
As a polymer chemist I was often also the rheologist (not common, I self-taught under some guidance from a thesis committee member) and the formulator and the applications tester. I was very comfortable with linear rheology, I even did a company-wide training once at Hexion, but anything that was non-linear I'd prefer a real rheologist to figure it out, but those are kind of few and far between at a chemical company.
I think this could be a post of its own, but most chemists right now I think are expected to handle more and more of the jobs of their co-workers. This might mean plant support (they fucked up and need you to figure out how to rework off-spec material), customer support (product isn't working correctly), process improvements (e.g., batch time from 24 hours to 8 hours), and various cost reduction projects. That's in addition to "figure out that next product that will bring in new revenue."
I think most chemists wish they could spend time coming up with new stuff, but there just usually isn't enough time in the day or resources at your disposal to do it. Also, plenty of "by-product" out there that could be gotten for cheap and turned into new polymers (BPA distillation bottoms, vanillin distillation bottoms, asphalt like stuff, etc)
Thanks! Figures that really you have to do a bit of everything.
Sounds like a big part, then, is turning those new by-products into current products that the company knows how to sell?
If you have access to them, I only gave it as an example because it was non-obvious, but I think you can also just view organic chemistry as a tool in the toolbox of the chemist too.
If you're into books and spreadsheets, I love Van Krevelen's 'group contribution' method for predicting properties of polymers but it's not as inspiring as a whole simulation.
The Molydyn UI needs some work. It's a bit too raw to invite a casual user and something close to click/drag interface like Chemdraw would be more approachable to a pure synthesis person looking to get started in simulation. Someone completely new isn't going to know what LAAMPS is or Leonard-Jones potentials, there has to be some hand holding and oversimplification with a JMol kind of input/output for most, and an Advanced Mode for experienced MD users who could enter all the parameters in just a DOS prompt if they wanted.
I should probably watch their tutorial haha. Even Gaussian for DFT or Abalone for MD are easier to hack around in and at least get something wiggling on the screen.
Thank you for the honest feedback! You're spot on with the need to simplify things - right now we're catering to computational chemists who've already got experience with LAMMPS. The update that's coming soon will be our first step into supporting all chemists. Do get in touch at hello@molydyn.com and I'd be happy to give you a demo! Thanks, Matt
I don't think 'able to be accessed' (free trial access) and 'accessibility' (user competency in workflow and interpretation) parse out to the same thing in the context of high level physical chemistry.
Shouldn't a practicing computational chemist already have a terminal to a corporate or university server or Amazon Web Service?