Big in Japan: IoT cryo-EM as a leap towards ultimate automation

Whenever I hear about Internet of Things (IoT), I think of smart fridges that have been touted as the future for at least two decades by now. But I don't know a single person who owns one. And to be frank, I'm not sure if I even want a connected fridge. Running low on bread or milk, I prefer visiting my local grocery store, where interactions with cashiers are about as social as I get these days. However, a vision of connected devices that I find much more compelling is a concept that Dr. Toshio Moriya and Dr. Yusuke Yamada from the KEK -Japan’s renowned High Energy Accelerator Research Organization - present in an AWS HPC Tech Short on YouTube. They propose to apply the idea of IoT, not to household appliances, but to cryo-EM:

To appreciate the transformative power of IoT in cryo-EM, it's crucial to understand the inherent challenges this field faces. Despite its incredible value, cryo-EM is far from being a black-box technology. It requires a high level of expertise and knowledge to utilize effectively, creating a steep learning curve for novices. Moreover, cryo-EM facilities, which require significant investments to be set up and maintained, are often under enormous pressure to deliver results and justify the resources spent on them.

The Catch-22 for cryo-EM facilities

The problem here is twofold. Firstly, the need to train new users is paramount to ensure the optimal and continued productivity of cryo-EM facilities. But ironically, the time and resources spent on training often impair the very productivity they need to uphold. Secondly, cryo-EM facilities often lack an influx of good internal projects and samples that can assure high success rates, further straining their efficiency. Meanwhile, many potential users confront high barriers to entry and face a scarcity of resources for learning and mastering cryo-EM techniques. In essence, the cryo-EM community is caught in a Catch-22 – a desire to expand the user base and projects, but a limited ability to invest in the required training and support.

Cryo-EM remains hard to learn, and even harder to teach

The journey from novice to proficient requires not only dedication but also a deep understanding of the entire workflow – from sample preparation to image processing. This knowledge isn't easily acquired through going through the motions of tutorials or visiting workshops; there's simply no substitute for hands-on experience with a real world project.
The bulk of the responsibility to shepherd new users through this complex process currently falls on the shoulders of facility managers. These managers are charged with guiding users through each stage of the workflow, providing the repetition and iteration needed to truly grasp the process. But, with each iteration taking a considerable amount of time, the managers can only support a small number of individuals adequately.
This situation is further exacerbated by the fleeting nature of successful trainees. Once individuals have mastered the techniques and generated meaningful results, they often disappear to prepare their findings for publication. As a result, the time and effort invested in training are in essence lost to the facility.

Processing computers are the first barrier for beginners

Embarking on a journey into the world of single particle cryo-EM inevitably leads to the foothills of a significant barrier: the processing computer. Mastering the cryo-EM workflow necessitates proficiency in processing given its integral role in the feedback loop with sample preparation and imaging. However, securing access to the right computing environment for cryo-EM is far from trivial for beginners.
Cryo-EM computing needs are intricate, requiring specific hardware and software environments, as well as robust data management strategies. The stakes are high – an inadequate setup can stifle the rapid, efficient feedback needed for good sample preparation and imaging.
A common issue faced by many novices is the lack of access to a high-performance computing cluster that's optimally configured for cryo-EM. And even when access to such a cluster is granted, it's not uncommon for it to be congested with jobs, leading to delays and inefficiencies.
An alternative to this is setting up personal cryo-EM workstations. However, this solution comes with its own set of challenges. Firstly, it requires upfront investment – a factor that may deter many. Secondly, the setup and maintenance of these workstations demand specialized knowledge, forming yet another entry barrier.
In essence, getting started with cryo-EM processing can feel like navigating a maze with walls made of technical requirements, hardware specifications, and software complexities.

Cloud removes the undifferentiated heavy lifting

The KEK team soon realized that the traditional model of setting up and maintaining individual workstations for a growing user base wasn't sustainable. The process was resource-intensive, required constant upkeep, and was particularly daunting for users without a deep understanding of the technicalities involved. This is where the transformative power of cloud computing stepped in.
In the cloud, building and maintaining a reference architecture for cryo-EM becomes a more manageable task. The cloud environment offers the flexibility to scale as per user requirements and eliminates the need for individual users to grapple with the complexities of setting up and maintaining their own processing units. Essentially, the cloud does the ‘undifferentiated heavy lifting' that was once an entry barrier for many potential users.
This way, users can focus their energies on the science, instead of being waylaid by the technicalities of cryo-EM IT setup and maintenance. The cloud approach provides an instantly accessible way for a broader base of users to engage with cryo-EM, bringing efficient cryo-EM processing to the fingertips of many more researchers.

A distributed system creates teachers out of learners

One of the most transformative aspects of the shift to cloud-based cryo-EM is how it changes the dynamics of learning and teaching. By making it easy and quick for beginners to get started, and scaling to accommodate a multitude of users who can share all details of processing, the cloud creates a significant critical mass of learners who progress in their knowledge together.
This shared learning experience fundamentally changes how problems are solved and questions are answered. Instead of having to constantly refer back to facility staff for every minor issue, learners in the cloud can lean on each other for support at first. They start helping each other, pooling their cumulative knowledge to overcome hurdles. This quickly transforms the learning experience, turning learners into teachers and fostering a sense of communal problem-solving.
As a result, the most experienced experts, who would usually be swamped with mundane technical issues, can focus their attention on the most challenging scientific problems. This shift frees up valuable resources, optimizes productivity, and most importantly, it creates a more efficient learning ecosystem where each user's progression benefits the entire community.

Japan turning unique predicament into inspirations for the global cryo-EM community

In retrospect, it doesn't surprise me that this transformative idea of IoT cryo-EM is born in Japan, a nation known for harmoniously blending a collective, collaborative culture with a strong aspiration for scientific leadership. Japan has always viewed itself as a powerhouse in structural biology, with a significant standing in fields like crystallography and nuclear magnetic resonance (NMR) as well as electron microscopy (Japan is home to two of three major EM instrument makers). Its scientific community excels in painstakingly working out the production of complex membrane proteins, a skill which typically provides a significant advantage in the field of cryo-EM.
However, with the advent of the "resolution revolution" in cryo-EM, Japan found itself trailing behind countries like the US, China, UK, and Germany. This unexpected position jolted many in the Japanese scientific community, given their high standards and inherent pride in their work.
Yet, in true Japanese spirit, this setback was viewed not as a failure, but as an opportunity for introspection and growth. And the drive to catch up and reclaim their position has been fierce. Already, the signs of this turnaround are evident, and the nation's scientists are making significant strides in cryo-EM.
But there is more to this story than national pride or the desire to excel. Japan's experience is an inspiration to other countries with scientific communities that feel left out from advancements in fields like cryo-EM. It demonstrates how, by fostering a culture of collaboration and sharing, we can achieve more with less. It reinforces the truth that in the realm of science, the whole is often greater than the sum of its parts, and that by working together, we can scale even the steepest of learning curves.

Connecting all available microscopes to a virtual central facility

Japan's main challenge, being short on high-end cryo-EM data collection tools, inspired a novel approach to their problem. Despite the shortage of top-tier microscopes, there was an abundance of low and mid-range ones scattered across the country. This lead to an inventive idea of connecting these scattered microscopes to a regional, or even a nationwide network. This ingenious solution is not merely interesting. But it fundamentally reimagines how learning and project development are approached in the field of cryo-EM.
In essence, this concept decentralizes the learning process and project development, allowing the entire scientific community to take part in cryo-EM. Once a project matures enough to warrant the scarce data collection time on high-end tools, it is prioritized, ensuring optimal resource allocation. The idea not only offers an elegant solution to Japan's unique situation but also presents a blueprint for communities in other countries facing similar constraints. 

IoT cryo-EM breaks down the walls of laboratories and promotes open science

Such a decentralized model, made possible by the concept of IoT cryo-EM, effectively dismantles the walls that traditionally separate laboratories. With these barriers removed, scientists are able to collaborate across affiliations, fostering a culture of cooperation and knowledge sharing. This profound change isn't just a shift in logistics or operations, but also in mindset. No longer are laboratories seen as isolated entities, but as nodes within a broader, interconnected network, contributing towards a shared scientific pursuit.
In this new paradigm, the lines between individual efforts blur, converging into a collective journey of discovery. The concept of IoT cryo-EM does more than just connecting microscopes or pooling resources. It actively promotes an open science approach by creating an environment where data, methodologies, and findings are shared openly.

Open science enables sharing of annotated data for ultimate automation

One of the defining features of an open science model is the ability to share data including annotated data freely across the community. Annotated data, in this context, refers to the user inputs and parameters used during the processing of their datasets. Right now, cryo-EM image analysis still requires substantial human intervention and expertise. But the ultimate aim is to make this process as black box as possible, reducing the complexity and knowledge required to a minimum. 
The collection and dissemination of detailed, user-specific data provides the entire community with invaluable insights into the nuanced decision-making processes and unique techniques applied by diverse researchers. More importantly, sharing annotated data has profound implications for automation in cryo-EM image analysis. Such datasets serve as a powerful learning material and knowledge base for refining algorithms, optimizing processes, and ultimately improving the accuracy and efficiency of automated workflows. When machine learning algorithms are fed with these annotated datasets, they can be trained to mimic the approaches and strategies employed by expert users, thereby moving us closer to the goal of making cryo-EM more accessible to the wider scientific community.

Structural biology becomes a mainstream method

With the achievement of ultimate automation, the field of structural biology is poised to evolve from a niche domain into a democratized research tool accessible to the broader scientific community. In this future scenario, cryo-EM will take its place as one of many common methods in the repertoire of molecular biologists, further empowering their research.
While some might feel a touch of nostalgia contemplating a future where the intricacies of today's structural biology workflows become streamlined, black-box processes, we should not shy away from this progress. The current state of cryo-EM, riddled with pesky non-intellectual challenges, is not a particularly desirable one. We should look forward to the day when we can bid farewell to these issues and focus our efforts on pushing the boundaries of our understanding of the biological world.
IoT cryo-EM will hasten our journey towards this future, opening up new possibilities and amplifying the impact of every researcher's contribution. In doing so, it will play a pivotal role in transforming cryo-EM from a niche, challenging area of study into a powerful tool accessible to researchers everywhere.

Useful links

1. The Challenges of CryoEM with our friends from KEK in Japan (Part 1 of 4)

2. KEK's novel solution for CryoEM's software and infra (Part 2 of 4)

3. How KEK changed how everyone in Japan does CryoEM (Part 3 of 4)

4. KEK's arsenal of CryoEM benchmark data - a detailed walk through (Part 4 of 4)