Iterative Mapping set to unlock new depths in multiomics

Parag Mallick and Sujal Patel co-founded Nautilus Biotechnology with the goal of harnessing protein analysis to generate high-resolution insights and uncover the full complexity of the proteome. The company recognizes that, while proteomics technologies are incredibly powerful, they haven't been nearly as widely used as genomics technologies and come with challenges in coverage, sensitivity, data quality, and ease of use.

“Rather than fix or improve existing methods, we wanted to create an entirely new proteomics method focused on sensitivity and scale without compromising other qualities,” shares Parag Mallick, Associate Professor at Stanford Medicine and co-founder and Chief Scientist at Nautilus Biotechnology. “In genomics, they achieve this sensitivity through amplification; they start with a single molecule of DNA and copy it. You can’t really do that with proteins, so we needed a single-molecule method.”

“Our next step was developing a way to identify individual molecules,” adds Mallick. “Most protein measurement technologies, such as ELISA or western blots, are a single measurement. We wanted to get as many measurements of each individual protein molecule as we could to enable greater resolving power.”

As a result, Nautilus began developing its Iterative Mapping method, which interrogates proteins at the single-molecule level and is easily scaled using large arrays capable of holding billions of proteins immobilized on individually distinguishable landing pads. In this novel method, every protein molecule is analyzed at the single-molecule level simultaneously through cycles of probing with fluorescently labeled affinity reagents, one reagent each cycle.

“Cycle after cycle, we're building up more and more detail about each individual molecule in this massively parallel way,” says Mallick. A machine learning framework then uses all the information to identify and quantify the proteins and proteoforms.

Working at detail and at scale

The Nautilus Proteome Analysis Platform uses Iterative Mapping for two key applications to start:

1. Targeted proteoform analysis, which uses very specific affinity reagents that recognize targeted sets of modifications and isoform-specific sequences, to analyze individual known proteins in great detail.

“The targeted mode looks at proteins that are post-translationally modified, and sees how this affects their behavior, which we can’t find out from the genome or transcriptome,” explains Mallick. “An example is our analysis of the protein tau, which is hyperphosphorylated in Alzheimer's disease, and is associated with disease progression. Understanding how different expression patterns of transcripts connect with tau and how these link with the disease are key questions that can only be answered in a multiomics context.”

2. Broadscale discovery proteomics, which uses affinity reagents that target very short epitopes within proteins to map the proteome more widely, revealing its breadth and quantifying proteins across a wide dynamic range.

“In broadscale mode, Iterative Mapping will shine where there are specific regulatory processes between the transcript and the protein scales, as its sensitivity allows researchers to observe very low protein levels, and spot differences in turnover kinetics for particular splice variants or isoforms,” explains Mallick. “As an example, in hypoxic shock, the cell suddenly turns off its degradation switch and the level of the protein HIF-1alpha skyrockets. Iterative Mapping would allow researchers to see the resulting transcript-protein discordance.”

In multiomics research, it is essential to have data that can be compared between different ‘omes. This makes it possible to see where there is agreement or discordance between different layers of biology and may point to mechanisms regulating that discordance that are targetable with therapeutics. Yet today, proteomics platforms cannot analyze proteins with enough detail to make these comparisons in many cases. Mallick explicitly set out to change that with Iterative Mapping.

“In contrast to mass spectrometry, which can have missing data because peptides don’t ionize very well or are masked by high-abundance peptides, all the Iterative Mapping measurements are independent,” states Mallick. “So, you will get confident identification and count-based quantification for both low and high abundance proteins. This makes the results readily integrable with RNA-Seq data.”

Where Iterative Mapping and AI meet

Mallick’s academic lab at Stanford is independent from Nautilus but works on projects that may be accelerated by Iterative Mapping and the Nautilus Platform. For instance, his lab is developing AI systems, including foundation models, which aggregate information across a huge number of datasets and can be used for generative or discriminative purposes. Mallick describes these as base models that coalesce as much knowledge as possible.

“These models are hugely sensitive to bias, and to inaccurate or poor data,” shares Mallick. “It has been shown that if as little as 1% of the data is misleading, it can actually distract the foundation model. Because of this, having incredibly accurate data becomes even more important. We believe that Iterative Mapping is going to generate data that is particularly well poised for training by or training with foundation models.”

A lot of the changes that researchers look for in disease are outlier or rare event effects. For the foundation model to find these changes, they need an ultra-sensitive platform, especially as missing data is really challenging for a foundation model to deal with.

The power of multiomics

Multiomics analyses show the interactions and connections between genomes, transcripts, proteins, metabolites and more while also showing how these layers of biology are regulated.

“One of the things that we need to keep in mind when we are carrying out multiomics analyses is understanding where the pinch point is in the process we’re studying. Sometimes that might be at the transcript level, sometimes that might be at the protein abundance level, sometimes that might be at the protein location level, or it might be a post-translational step,” concludes Mallick. “When you are looking at multiomics, you might actually see no change to the transcript or the protein level, because all of the critical events are happening at the post-translational level. When we can get to the point that we can look for the critical decision points in a process, that's when multiomics will really be able to fulfill its promise.”

Nautilus is developing the Nautilus Platform with the goal of making it possible to more efficiently identify those pinch-points and find new ways to monitor and drug biological processes that are intractable today. In doing so, they hope the vastly accelerate multiomics research and enable scientists to leverage multiomic discoveries to guide the development of next-generation biomarkers, diagnostics, and precision medicines.

To hear from Parag, Nautilus, and Birgit Schilling, an early user of the Nautilus Platform and Professor as well as Director of the Mass Spectrometry Core at The Buck Institute for Research on Aging, register for their upcoming webinar.