Faster, More Confident Small Molecule Structure Verification in Pharmaceutical R&D
Explore how ASV transforms structure verification into a scalable, consistent and reliable process.
Complete the form below to unlock access to ALL audio articles.
Structure verification in R&D
Chemical structure verification is a tedious but critical task in the drug discovery and development pipeline. Despite the advances in automation, AI and high-throughput screening that have revolutionized many other workflows in pharmaceutical R&D, structure verification continues to be held back by the tug of war between speed and accuracy.
Structure verification relies heavily on nuclear magnetic resonance (NMR) data. Manual interpretation of NMR data is slow and requires significant expertise. The challenge becomes even greater at the scale and speed of modern pharma R&D, where that level of expertise isn’t always available.
At the same time, accuracy of chemical structures underpins virtually every scientific and regulatory decision made across a molecule’s lifecycle – from hit confirmation to regulatory submission. Structural errors or ambiguities can waste resources, derail programs and delay development.
As a result, scientists in pharmaceutical R&D and beyond continue to search for ways to lessen – or better yet, eliminate – the bottleneck of structure verification without compromising accuracy.
The case for automated structure verification
Automated structure verification (ASV) tools compare proposed chemical structures to raw experimental NMR spectra using various algorithms and rules. The general principle is that the spectra of the proposed structure are predicted using software and the prediction results are compared to the experimental data.
Some metrics are calculated in the process, like the match factor (MF). These metrics detect inconsistencies, errors or ambiguities in the proposed structures. This verification process saves scientists time and reduces the risk of human bias, as analysts will subconsciously try to fit the data to the expected structure.
Several pharmaceutical and chemical organizations have now deployed ASV systems as part of their structure verification workflows. Stories from scientists who are using these systems across various stages of R&D demonstrate ASV’s ability to boost modern workflows and pave the way towards fully autonomous structure verification.
ASV accelerates structure verification throughout R&D
In development: Freeing scientists to tackle complex problems
Scientists in pharmaceutical development analyze fewer compounds than their colleagues in discovery but often face more structural ambiguity.
That was the case for Novartis’s Chemical Structure Investigation (CSI) team, who has been tasked with elucidating the structures of compounds in development and commercialization. For many of the compounds that the team deals with, elucidation can take weeks, making rapid verification of a suspected known structure desirable. ASV helps make this a reality.
The CSI team estimates that experienced analysts typically spend 20 minutes on manual NMR verification per submission. However, when ASV delivers a result, review takes just 1–2 minutes – up to 90% time saved. If small manual corrections are needed, the process still takes only 5 minutes. If ASV doesn’t return a result or the data is incomplete, the pre-processing and referencing of the spectra still saves time.
Figure 1: Approximate expert time required to analyze one submission with and without ASV. Credit: Image courtesy of Novartis.
In discovery: reducing time to lead optimization with high-throughput NMR
Chemists in drug discovery, on the other hand, often handle large numbers of compounds and have a good idea of the expected structure – whether from design software, reaction predictions or supplier metadata. In these environments, the overall time savings from ASV quickly add up.
Verifying structures of many small molecule building blocks is a burden familiar to scientists in the early stages of R&D in the pharmaceutical and agrochemical industries. When Syngenta undertook the quality control of the more than 2000 samples in its chemical stores, the team compared the time it took to manually analyze the NMR data of a batch of 96 samples to that with ASV. They found that the use of an ASV system cut down the amount of expert analysis time by almost a half.
Figure 2: Time required to analyze a batch of 96 samples with and without ASV. Credit: Image courtesy of Syngenta.
In the pharmaceutical space, Amgen performs quality control for compounds in the sample bank or commercial/externally synthesized compound libraries in discovery. Enabled by ASV, its high-throughput NMR workflow analyzes 25 compounds per day, totalling 5000 compounds annually. For a 96-sample well plate, results are ready in just 5 business days.
ASV enables more confident decisions
Increase in speed is not the only benefit ASV offers. Because it is free of human bias and error, ASV also allows scientists to verify structures more confidently.
The key to accurate ASV is to minimize false positives. This is because false positives are more likely to let erroneous structures slip past undetected and mislead downstream teams, whereas false negatives are easily identified in the manual review of the negative or ambiguous ASV results.
A well-calibrated ASV system can avoid most false positives while keeping false negatives low enough to maintain substantial time savings – even in environments with many diverse structures.
How To Detect Trace Metals During Semiconductor Manufacturing
This application note demonstrates how advanced multi-quadrupole ICP-MS technology can revolutionize ultra-pure NMP analysis for effective semiconductor manufacturing quality control.View App Note / Case Study
At Novartis, ASV is integrated into the open-access NMR system used by more than 400 users from different teams across R&D. It handles samples varying in properties such as molecular size, exchange phenomena and concentration.
In a one-day snapshot, 96% of samples had experimental data that were applicable to Novartis’s ASV system. The structure was accurately verified (true positive) or rejected (true negative) for 40% and 15% of samples, respectively. Another 39% of samples were wrongly identified as having the incorrect structure proposed (false negative) while only 2% were wrongly verified as the correct structure (false positive). Further investigation revealed that the false positive results originated from user error related to the sample and selection of experiments (i.e., the experiments selected were not able to distinguish the wrong proposed structure from the correct one).
- 116 samples
- Various experiments – from simple 1H to comprehensive 2D sets
- Wide variety of molecules
o MW 126–1168 Da
o Heteroatoms: F, P
o Different heterocycles
Figure 3: ASV results from open access NMR analyses over the course of a single day. Credit: Image courtesy of Novartis.
Train the algorithm for more impactful ASV
Even though ASV systems with accuracy as low as 50% can still provide substantial efficiency and confidence benefits, improving accuracy (i.e., reducing the number of false negatives without raising false positives) can boost the impact of an ASV system.
Advanced Solutions for Mining and Metallurgy
This guide highlights a complete portfolio of spectroscopy, chromatography and mass spectrometry solutions that support efficient analysis at each step, from exploration to refining.View How To Guide
Improving the accuracy of the system largely comes down to improving the accuracy of the underlying NMR prediction algorithm. While commercial prediction tools are more accurate than ever, they often struggle in specialized or proprietary chemical spaces. Even the best out-of-the-box predictors can fall short in these niche areas. Training prediction algorithms with experimental NMR data is a quick and easy way to enhance their accuracy.
Sanofi has shown that by adding as few as one related structure to the prediction training database, the accuracy of the predicted spectrum, and thus the MF resulting from ASV increased significantly.
Figure 4: Spectral prediction – and thus verification – accuracy increases significantly as more relevant structures are added to the prediction database. Credit: Image courtesy of Sanofi.
Laying the groundwork for greater efficiency gains
As more chemistry becomes automated in pharmaceutical R&D, the amount of data chemists must deal with continues to increase. To alleviate some of this burden, AstraZeneca is working towards a fully autonomous structure verification system.
The goal is a system that needs minimal NMR data and still distinguishes isomeric reaction products. Occasional failures are acceptable – if the system “knows when it doesn’t know” and flags for manual review.
With ASV playing a foundational role, AstraZeneca is exploring the incorporation of reaction product structures proposed by generative AI, as well as the use of data from other analytical techniques.
Structure verification by NMR is a vital task in drug discovery and development. Done manually, it is slow, subjective and incompatible with the speed and scale of modern science. ASV transforms it into a scalable, consistent and reliable process.
ASV is being used by major pharmaceutical and chemical organizations to accelerate the R&D pipeline and prevent costly mistakes. As organizations look toward fully integrated and autonomous R&D environments, ASV will be a cornerstone technology for ensuring that every decision – from hit validation to regulatory submission – is built on solid structural ground.
