Spectronaut™, our flagship software recently turned 10 years old. Over the past decade, Spectronaut has improved greatly in its functionality and interface and has earned recognition as the gold standard for DIA analysis.
From the start, we have placed emphasis on continuously developing Spectronaut to meet the needs of the proteomics community in general and our users in particular. We have no plans to slow down, and our developers are adding the finishing touches to Spectronaut 16 which will launch in June 2022.
Here, we reflect on how Spectronaut has developed through the years, where it stands today, and what improvements are to come.
As many software solutions do, Spectronaut started out with a humble interface and limited functionality. The first version of the software was coded by Oliver Bernhardt, an intern at Biognosys at the time, now a Principal Scientist in our Bioinformatics team and the lead developer of Spectronaut.
In 2011, for the first version of Spectronaut, we focused on creating the core aspects of the software, making sure that its functionalities could develop and grow for years to come. Back then, Spectronaut already allowed the user to look at the XICs, a feature that was, and still is, unique.
Figure 1: Spectronaut 1 included the core aspects seen in Spectronaut today.
From the early stages, we recognized that having a user-friendly interface was crucial for the overall user experience of our software, this became a focal point for the release of Spectronaut 2. In the second version of the software, the user interface was significantly improved with the addition of quick access to perspectives and an improved graphical experience, shown below.
Figure 2: In Spectronaut 2 we greatly improved the user experience of the software.
Today, Spectronaut is a software that combines a seamless user experience and functionality in a unique way. This, coupled with our dedicated customer support makes Spectronaut the ideal companion for any DIA project.
Figure 3: The latest version of Spectronaut offers a seamless user experience.
Over the years, Biognosys has tested uncountable ways of acquiring and analyzing DIA data and we have made huge improvements in what can be achieved with single-shot DIA profiling.
The first time we acquired DIA data on an orbitrap platform was back in 2011. What we could achieve then with HeLa single shot DIA was no better than what could be achieved with DDA. When we made our first public release of Spectronaut in 2013, however, we had already made significant improvements. Since then, the improvement in both acquisition method and data extraction has been steadily increasing, resulting in a subsequent increase in identification performance, as shown below.
Figure 4: Improvements in the DIA acquisition over the years.
In 2017 we released Pulsar, our search algorithm, now integrated into all Biognosys software. With Pulsar, we enabled two key solutions. Firstly, searching data for library generation became native in Spectronaut, providing an all-in-one, seamless experience for DIA analysis. Secondly, Pulsar can search DIA data against a FASTA file. This led to the introduction of directDIA™, our library-free workflow that eliminated the need to acquire additional DDA data for library generation.
The iRT concept, invented by Biognosys, has been a key aspect of our DIA data extraction algorithms that has evolved through the years. The first versions of Spectronaut relied on a basic linear retention time calibration, supported by the use of our iRT Kit. With the release of Spectronaut 7 in 2015, we introduced our high precision iRT concept, another important step in the development of Spectronaut. With high precision iRT, it became possible to zoom in on ion chromatograms to identify and focus on real peptide peaks. This not only improves sensitivity but also speeds up the analysis making data processing faster and more efficient.
With Spectronaut 13 in 2019, the iRT calibration made an unprecedented development: deep learning assisted high precision iRT calibration. With the implementation of deep learning algorithms for iRT calibration, Spectronaut can predict iRT values, eliminating the limitation of needing empirical information for iRT in the dataset. This significantly increases the variability of datasets that can benefit from an optimal iRT calibration while improving the sensitivity and efficiency of the analysis.
Spectronaut has continuously become more efficient with significant speed and memory improvements. As shown in the image below, the analysis time has been reduced from 40 minutes to 10 minutes for the same data set when Spectronaut 5 and Spectronaut 15 are compared. We have also consistently decreased the memory requirements of the software and Spectronaut increasingly uses resources more efficiently. It is worth noting that these improvements did not compromise the richness of the results.
Figure 5: A comparison of speed and memory consumption between Spectronaut 5 and Spectronaut 15.
As previously mentioned, an important milestone in Spectronaut’s evolution was the introduction of directDIA™: the ability to search DIA data directly, thanks to the development of Pulsar. This workflow was a true game-changer for DIA proteomics as it eliminates the need for DDA-generated libraries and instead, directly searchers your DIA runs. This method has improved significantly in recent years and continues to do so.
directDIA is performed in two steps. First, DIA data is extracted into MS2 spectra that are then searched using Pulsar, Spectronaut’s integrated search engine. In the second step, the search results are used to generate a spectral librarythat is used in the targeted analysis of DIA data. This workflow results in a high-quality and consistent quantification, minimizing missing values. Because directDIA utilizes a library generated from the experimental DIA runs, this approach provides the best results when applied to real, biologically relevant experiments with a proper number of replicates.
Figure 6: The directDIA workflow.
With our second release of directDIA with Spectronaut 14 in 2020, we already reached a point where we essentially had the same performance with directDIA and deep project-specific libraries compiled with DDA data. This held true even in projects with high variabilities, including plasma studies with large cohort sizes. Today directDIA is highly quantitative and provides high-quality results that outperform those from project-specific libraries from DDA runs.
One of our priorities when developing Spectronaut is to provide the community with the best solution for DIA data analysis, independently of the MS platform used to generate the data. We have teamed up with different instrument vendors as well as with some of the most prominent method developers to ensure that we provide support for the latest innovations in data acquisition. As a result, Spectronaut is the most versatile software when it comes to workflow support.
The most recent example of this is our support for ion mobility, introduced in 2020 with Spectronaut 14. This introduction resulted in a very significant improvement in the number of identifications, allowing us to identify well over 200,000 precursors in our HeLa 2h single shot DIA run (see figure 4), something that was unthinkable a few years ago.
We continue to work and collaborate with method development experts to fulfill the need for support for the latest innovations in mass spectrometry.
Machine learning has been emphasized from the beginning stages of Spectronaut’s development but our first forays into deep learning started in 2016 when we first looked at retention time prediction using deep learning. We quickly saw great potential for applying deep learning in the proteomics world and have continuously developed this feature ever since.
We have focused heavily on integrating deep learning into Spectronaut and were among the first to release a commercial software with deep learning that made active use of it. The deep learning feature has continuously improved and can now predict fragmentation, ion mobility, and iRT.
Going forward, we have identified numerous possibilities of using deep learning within Spectronaut and look forward to developing this feature even further.
Spectronaut 16 will bring significant improvements, including changes to our peptide-centric way of data analysis and deep learning. Our focus is on improving both identification and quantification as well as continuing to support users working with PTMs and further improving directDIA.
Beyond Spectronaut 16 we look forward to future improvements that will further increase our ability to extract data from convoluted DIA runs while also providing even better quantification. Alongside these improvements, we will continue improving the user experience for many aspects of the DIA analysis journey.