Looking Towards a Digital Future for Analytical Chemistry Post-COVID-19

The Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy (Pittcon) is, in non-pandemic years, the world’s largest conference for laboratory science. In 2021, Pittcon has moved online, mirroring the remote and disrupted workplaces of most of its attendees. We spoke to regular Pittcon exhibitor ACD/Labs, which specializes in informatics platforms for analytical data handling. Andrew Anderson, ACD/Labs’ vice president for innovation and informatics strategy, outlined the challenges analytical chemists have faced during the pandemic and how informatics tools can help to make digital strategies, both remote  and in the lab, more streamlined.

Ruairi Mackenzie (RM): You spoke at last week’s Pittcon – could you please summarize the key points of your presentation?

Andrew Anderson (AA): The pandemic and the corresponding travel restrictions have presented lots of challenges for scientists. This is my first virtual conference, and I think it's emblematic of the larger challenge of digital experiences.

Over the last year or so, we've really seen in an uplift in demand for our tools and products that allow for more interactive access to analytical data. My focus at Pittcon this year was, at a high level, about digital contextualization of the work that goes into characterizing materials. That characterization process is not binary; it’s not “uncharacterized versus fully characterized”. There's a lifecycle that is linked to the effort of comprehensively characterizing the composition of a substance. A lot of our customers use analytical instrumentation to characterize drug products. That characterization process starts with techniques like chromatography, which leads, ultimately, to a mass balance, where users can calculate impurity profiles and the composition of the substance at a very accurate and precise level, so that every unit of mass is associated to different compositional components.

In the beginning, users may know what the composition is – they can see the peak in the chromatogram, but the identity of every individual component, such as impurities or degradants, may remain unknown. What I presented on is how companies should build their digital systems to not just account for the final stage of fully characterized substances, but for the whole lifecycle.

The journey to go from those initial chromatograms to the more fully described mass balance is largely predicated on the type of data accessed, how the data was stored and managed, and how it was annotated, so that as more information about a material or component’s identity is acquired, it is fully accounted for in the digital structure.

I then discussed some of the digital tools we have available in this area. One is the ability to take, for example, raw chromatographic data or processed chromatographic data that contains detected peaks, integral values for those individual peaks and corresponding attributes like asymmetry or tailing factors and provide identity to individual peaks.

 At the outset, users may find that the only identity they have is a retention time value. That's a very method-specific identity. Then, let's say they applied their scientific knowledge and gave a candidate identity for a peak – if the LC-MS data suggests that a particular peak is an impurity related to the API, it could be, for example, a desmethyl compound.

Once they have assigned that first candidate identity, our software allows for facile structure elucidation of substances. Over the lifecycle, as they acquire more data, users will be able to give more precise chemical structure identities to a substance that are independent of the analytical methods. The software and digital systems that manage that lifecycle have to be able to handle processed data and then associate it with what I’d call interpretation results; in our example, that initial interpretation, identifying desmethyl in the API.

The final component that has to be included in that digital strategy is what I'd call chemical structure representations associated to those individual components; peak tables as an example. What I was trying to make clear to the community is that having the ability to represent that data-to-identity lifecycle in their digital structure is paramount to being efficient. And at the same time, really mitigates risk. For example, they may see a peak in a drug substance impurity profile. If they have this lifecycle accounted for in their digital structure, the next time they see a similar peak in another sample, they can easily make an association between chromatographic information, like peak retention time and peak shape. Therefore, they are able to digitally assign that previous identity to the new lot or batch of material that they are currently testing, so it streamlines the dereplication process. When new lots of drug substances come in and the user is relegated to performing that compositional assessment, they can now rely on their own legacy data to streamline that process.

RM: What else did ACD/Labs focus on at Pittcon?

AA: A trend we have seen in general among our customers is a strategic interest in digital transformation. The journey I just mentioned, the identity lifecycle of a composition, is a key facet of digital transformation.

What we've seen is investment and interest in digitalizing and moving to the cloud. What we've been working on with our customers is augmenting their cloud and Dx strategies to account for analytical data. So most recently Pfizer and ACD/Labs co-presented Pfizer’s new digital strategy, explaining how they're leveraging our Spectrus platform to enhance the cloud experience for their scientists.

As an example, they are leveraging our unique capabilities in making cloud data available on demand. The prerequisite there is how to know exactly what you are looking for on the cloud. Without the proper search tools, looking for analytical data in the cloud is like looking for a needle in a stack of needles, because IT organizations will often focus on the marshalling of data from the source to the cloud. If the user knows what they are looking for, let's say the data were organized by file type and they have the file name, they have to hope that the metadata is exposed for that individual data file and they can query it.

But often scientists don’t think to look for metadata, rather they may look for data features. For chromatography data, this may be features such as retention time or retention time plus integral value.

These are types of searches that, practically speaking, ACD/Labs can absolutely provide. If the user wants to search by spectrum, our software allows them to search for all spectra that have, for example, three peaks at three particular values. What that gives users is an additional ability to distinguish needles from needles, enhancing the cloud experiences to provide we might call domain-specific query types that let users search by:

• Chemical structure
• Analytical feature, e.g. chromatography feature or spectral feature
• Metadata

Pfizer layered our Spectrus platform onto the cloud, adding a querying layer that allowed for more facile query capabilities and on-demand data provision for their scientists when they needed it, but in a language that they speak; giving them the ability to use scientifically relevant query terms like structure or spectrum or chromatogram as opposed to purely text and numerical searching. 

Andrew Anderson was speaking to Ruairi J Mackenzie, Senior Science Writer for Technology Networks