QuantumFuture Scientific

From Lightning-Fast DFT to Precision Docking

QF Docking Methods

QuantumFuture (QF) now offers docking services powered by our cutting-edge quantum-enhanced docking methods. These innovative approaches integrate precomputed high-quality conformational libraries, featuring QM-optimized geometries and ab initio DFT-D4-based conformational rankings and strain energies.
Our methods significantly enhance docking-based virtual screening accuracy by resolving critical issues in intramolecular relative energies—errors that plague traditional flexible ligand docking approaches. Extensive research demonstrates that widely used force-field-based methods introduce substantial inaccuracies in conformational relative energies. Consequently, the scoring functions typically employed in docking cannot be expected to deliver accuracy.
Traditional flexible ligand docking often generates high-strain geometries with energy deviations comparable to protein-ligand binding energies—far exceeding the differences used to prioritize hits in virtual screening. By incorporating physically realistic ligand conformations with rigorously computed strain energies, QF docking yield markedly different and more reliable docking outcomes, even when using the same protein-ligand interaction scoring functions.
We have already applied these methods to real-world challenges, including COVID-19 protein targets, and observed significant improvements in docking hitlists and predicted protein-ligand interactions.
QuantumFuture’s most important docking service is based on our unique collection of 30,000 natural products that are estimated to be safer than the least toxic 30 percent of all approved drugs. Toxicity remains one of the biggest barriers in drug development, often leading to failures in clinical trials. 
Our approach stands out because traditional docking methods suffer from unreliable conformational energy estimates due to inaccuracies in docking scoring functions. Our methods integrate high-quality quantum-mechanical conformations, DFT-D4 strain energies, and accurate scoring functions to ensure that only the most physically realistic and stable ligand geometries are considered. This leads to better accuracy in hitlist ranking.

QuantumFuture offers flexible collaboration models and service tiers for academic groups, biotech startups, and pharmaceutical companies. 
We are actively seeking collaborations with research institutions, biotech startups, and pharmaceutical companies to advance precision drug design. If you are interested in discovering highly potent yet ultra-safe drug candidates, contact us today to get started

IntroductionPaper (at free online journal)Podcast (NotebookLM)Short Video (6-7 minutes)

Fighting Cancer and Alzheimer’s Disease: QuantumFuture’s Docking (QFD) Identifies Dozens of Potent, Safe IDO1 Inhibitors from Natural Products

Indoleamine 2,3-dioxygenase 1 (IDO1) is implicated in Alzheimer’s disease, Parkinson’s disease, and various cancers, underscoring the need for potent and nontoxic inhibitors. In our previous work, we introduced QFVina and QFVinardo, highlighting the integration of high-level quantum mechanical (QM) conformational data with AutoDock Vina’s vina and Vinardo scoring functions. Building on these methods, we now introduce the generalized QuantumFuture’s Docking (QFD) framework and demonstrate its first broad application, featuring a newly curated library of approximately 30,000 natural products, each rigorously screened for very low predicted toxicity. Using ab initio DFT-D4 calculations (rev-SCAN functional, def2-TZVP basis set) to obtain accurate conformational strain energies, QFD was applied to 16 protein targets (eight with and eight without a heme cofactor), revealing several dozen top-scoring ligands that combine strong binding with minimal toxicity. Interestingly, the Vinardo and dkoes scoring functions emphasized different subsets of promising hits, highlighting the advantage of using multiple scoring perspectives. These potential IDO1 inhibitors are prime candidates for further refinement via QM/MM and/or molecular dynamics (MD) simulations to account for ligand and protein flexibility. We encourage experimental validation to accelerate the development of these low-toxicity compounds for therapeutic applications in neurodegenerative diseases and cancer.

Paper (free journal)Podcast (NotebookLM)Short Video (6-7 minutes)

Ab Initio DFT Developments

We are developing an extremely efficient linear scaling DFT program with the advanced QFC (Quantum Future Coulomb) method for linear scaling, fast and accurate Coulomb interactions and a new atomic grid based technology for exchange-correlations (QFXC). Unlike other similar programs our idea is to keep the full accuracy of the traditional DFT calculations by using Gaussian basis sets. There is a large history of using such ab initio programs for many decades now and an unbelievable amount of experience has been collected in this field. Keeping the full accuracy of those methods makes our job, as well as the job for our users easier since we do not have to prove that the results are useful and can be trusted. All we have to show is that it is much faster than alternative programs and provides essentially the same results. Ab initio DFT programs are widely and routinely used today in many area of drug design and, especially on public clouds, research groups often invest hundreds of thousands or even millions of dollars in such calculations. We would like to invite those research groups to consider reducing a million dollar ab initio DFT calculations project to less than a hundred thousand, save 90-95% of money, save over 90-95% of carbon footprints and have the same results at the end. Although our program needs licensing and is not free it is much cheaper than “free” once we consider that computation times are not free and our power sources are not really the outlets on the wall either.

DFT Speed BenchmarksDFT-D4 Accuracy BenchmarksQF2024 Release NotesLarge Project with Ease on AzureShort Video (6-7 minutes)

QM Based Conformation Search and Torsion Scan

There are numerous commercial as well as freely available conformation search and torsion scan programs available today. We offer a much more robust alternative with QM quality scoring. There are many projects in computational drug design where one of the first steps is to look for low energy conformations of the given ligand, and then a large amount of computational resources are spent in later steps using those conformations. Obviously, we do not want to choose conformation search programs that provide results within a few seconds or, at most, a few minutes but may risk overlooking very important and sometimes even the most critical low-energy conformations because all expensive computational steps followed the conformation search would be wasted, or the results would be wrong or misleading. This is why we offer a slower but more robust alternative.

Benchmarking with FDA drugsDrug Repurposing Conf. DBLarge Project with Ease on AzureQF2024 Release Notes

QuantumFuture’s Scientific Database for Drug Repurposing

1. Extensive Conformational Library: Explore low-energy conformations of approximately 1600 FDA-approved drugs and about 3900 natural
compounds. All conformations are ranked with accurate DFT-D4
revSCAN, def2-TZVP ab initio calculations.
2. Thermodynamic Insights: Access comprehensive quantum mechanical thermodynamic calculations for each drug molecule’s conformations as well as for the conformational ensembles.
3. The database can be used in various ways in drug repurposing projects. One example is screening for new protein targets with a specialized docking program, utilizing the precomputed and physically sound conformations, and considering accurate ab initio DFT-D4 based
conformation strain energies in the docking hit selection processes.
4. Accessibility and Appreciation: Enjoy complimentary access for special QF customers and Microsoft-affiliated research groups, courtesy of Azure credits. A nominal fee is charged for others to help cover our operation expenses.
5. 2024 Software Release: Unlock new drug possibilities with our 2024 QF software, capable of handling similar projects involving tens of thousands or even hundreds of thousands of private drug-like compounds.
6. Project Outsourcing: For those seeking to outsource similar projects, we offer our expertise and capacity to meet your research needs.

QF Drug Repurposing Conf. DBBenchmarks with FDA Approved Drugs

Protein Ligand Interactions

Looking for specific molecules that efficiently inhibits a given protein target but does not interfere with the functionality of other essential proteins is a very difficult and complicated process. The sampling requirements are very large, lead molecules together with the given target protein and water molecules form very complex systems with thousands and tens of thousands of degrees of freedom and for this reason the majority of the predictions are based on MD simulations today. These simulations utilizing force fields that are proven to be very inaccurate in simple intermolecular interactions in vacuum or in molecular crystals where much more accurate solutions are already available and the accuracy can be validated. We believe that ab initio based DFT algorithms will play a very significant role in this research in the near future and our most important priorities are to be part of this very important journey. Stay tuned!

COMING SOON

Organic Crystal Structure Predictions

It is extremely important for pharma companies to find potential polymorph crystal structures (all possible low energy crystal structures) in order to avoid unexpected surprises later on that would costs billions of dollars to correct (it has happened already). Possible crystal structures and their energies are also essential for solubility predictions! It has been shown over the last 3 decades via blind challenges of crystal structure predictions that the only reliable methods are based on accurate large basis ab initio calculations with accurate dispersion corrections. It is also important to note that traditional force field methods based on atomic point charges (this is what usual MD programs utilize) have found to be absolutely unreliable since the difference in energy between potentially feasible crystal structures is much smaller than the error in those force fields. For these reasons it is very important to develop a very efficient ab initio DFT method that preserves full accuracy and at the same time reduces the very large computational expenses. It is also important to develop a reasonably accurate force field (or AI based method) for the purpose of global sampling, because although the final results should be obtained by accurate DFT calculations the large amount of sampling for pre-screenings requires to utilize much faster crystal energy evaluations.

COMING SOON

Force Field and AI based developments

Developing force fields in order to try to mimic real physical inter and intramolecular interactions with fast model potentials has been going on since the beginning of molecular modelling and simulations by scientists with infinite optimism and sometimes even belief of miracles (many of them are moving to the AI developments nowadays). Despite their limited accuracy these methods play a very important role in computational drug design since their calculation efficiency is absolutely necessary for sampling purposes and molecular dynamics calculations. We have some scientific experience in this field by developing an advanced intermolecular force field utilizing ab initio DFT based atomic distributed multipoles that has been used successfully at the organic crystal prediction challenge in 2016. The disadvantage of such a scheme is that ab initio DFT calculations are needed in order to obtain the atomic distributed multipoles. While our new very efficient DFT program offers very significant speed ups it is still way too expensive compared to force field calculations and consequently it is still the bottleneck of such method. It is possible to keep the atomic multipoles and just transform them based on the geometry changes but that approach is highly approximate, especially with large changes in the molecular geometries. One solution of this problem in developing an AI based scheme in order to obtain the atomic multipoles and use this AI code to reduce the number of necessary ab initio DFT calculations to obtain the new multipole set when it seems necessary. An alternative solution is to optimize a more traditional force field but without transferability requirement and customize it for a small set of molecules or crystal structures. There are some programs exist like this already but we believe that our highly efficient ab initio DFT program provides big boosts in projects like that since the computational expense is the bottleneck in all of such solutions.

COMING SOON