🧬 biology

RAcolony: A Multimodal Platform Integrating Raman Spectroscopy and Automated Imaging for Efficient Construction of Personalized Gut Microbiome Biobanks

The authors present RAcolony, an intelligent, high-throughput platform that integrates Raman spectroscopy with automated imaging to efficiently construct personalized gut microbiome biobanks by significantly reducing redundancy and capturing high-resolution microbial diversity, including novel taxa.

Original authors: Yang Yan, He Jiang, Jian-Bo Ma, Min-Zhi Jiang, Chen Chen, Zi-Nuo An, Xin-Yao Li, Guozhong Chen, Xiao-Yang Zhu, Tian-Yu Li, Yulin Wang, Chang Liu, An-Qi Zhao, Xiao-Jia Han, Dalei Wu, Bei Li, Shuang-Jia

Published 2026-07-16✓ Author reviewed ⓘ

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yang Yan, He Jiang, Jian-Bo Ma, Min-Zhi Jiang, Chen Chen, Zi-Nuo An, Xin-Yao Li, Guozhong Chen, Xiao-Yang Zhu, Tian-Yu Li, Yulin Wang, Chang Liu, An-Qi Zhao, Xiao-Jia Han, Dalei Wu, Bei Li, Shuang-Jiang Liu

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

The Invisible City Inside You

Imagine your body is a bustling metropolis, and the gut is its most crowded, vibrant downtown. Inside this district lives a massive community of tiny residents: bacteria. These aren't just passive tenants; they are the city's engineers, chefs, and security guards, helping you digest food, fight off invaders, and even talk to your brain. Scientists call this community the "gut microbiome." For a long time, we knew this city existed, but we couldn't really meet the residents. We could see the skyline from space (using DNA tests), but we couldn't knock on their doors and say hello.

To truly understand how this city works, or to fix it when things go wrong (like when a bad neighborhood takes over), scientists need to grow these bacteria in the lab. This is called "culturing." But here's the problem: the gut is a chaotic place. If you try to grow these bacteria on a petri dish, you end up with millions of tiny dots (colonies). Most of them look exactly the same, even if they are different species. It's like trying to find a specific person in a crowd of a million identical twins. The old way of doing this was to pick dots one by one with a tiny needle, hoping you didn't pick the same twin twice. It was slow, boring, and you often wasted hours picking the same bacteria over and over again, missing the rare, interesting ones hiding in plain sight.

Meet RAcolony: The Super-Smart Robot Detective

Enter RAcolony, a new high-tech platform designed to solve this "twin problem" and build a library of these gut residents. Think of RAcolony as a super-smart robot detective that can look at a petri dish and instantly know exactly which bacteria it is looking at, without ever touching it or killing it.

Here's how it works: The robot has two special eyes. The first eye is a high-definition camera that takes a super-clear picture of every single bacterial dot (colony) on the plate. It measures the shape, size, and texture—kind of like checking if a person is tall, short, or has curly hair. But the robot knows that two different bacteria can look identical, just like two different people can wear the same outfit.

So, the robot uses its second eye: a Raman spectrometer. This is like a magical scanner that reads the "chemical fingerprint" of the bacteria. Every type of bacteria has a unique internal chemical recipe (proteins, fats, and DNA) that creates a specific pattern of light when scanned. Even if two bacteria look exactly the same to the camera, their chemical fingerprints will be different. By combining the photo with the fingerprint, RAcolony can tell the difference between identical-looking twins instantly.

The Big Discovery: Less Picking, More Finding

The researchers tested this new robot on samples from healthy people. They compared RAcolony to the old-school method where humans picked colonies by hand, and a random method where they just picked dots without thinking.

The results were impressive. When humans picked colonies, they ended up picking the same bacteria over and over again, creating a lot of "redundancy" (waste). The robot, however, was much smarter. It managed to pick a diverse group of bacteria while picking far fewer total colonies. In fact, the robot achieved over 90% coverage of the different types of bacteria present on the culture plates (representing the bacteria that successfully grew), while keeping the "waste" (redundancy) down to about 150%. In contrast, the old manual methods often had redundancy rates exceeding 1000%, meaning they were picking the same bacteria many times for every new one they found.

But the robot didn't just stop at finding different species; it could even tell the difference between different strains of the same species. Imagine finding two people who both look like "John Smith." The robot could tell you that one is "John Smith from the bakery" and the other is "John Smith from the library" based on their chemical fingerprints. In the study, the robot found that even bacteria that looked the same often had different genetic makeups, suggesting they might do different jobs in the gut.

Building a Personal Library

Using this technology, the team built 8 personalized "Gut Microbiome Biobanks." Think of these as custom libraries containing the specific bacterial strains found in individual people. From just a few samples, they successfully isolated 167 different bacterial species, including 27 that might be brand new to science (potential new taxa).

Crucially, when they used four different types of food (growth media) to help the bacteria grow, the robot was able to capture 80% of the species found in the original stool sample. This is a huge jump from traditional methods, which often leave 50-70% of gut bacteria uncultured and unknown.

Why This Matters

The paper suggests that this technology is a major step forward. It doesn't just save time; it changes the game. By being able to quickly and accurately pick the right bacteria without wasting effort on duplicates, scientists can build better libraries of gut microbes. This is crucial for the future of medicine, particularly for things like Fecal Microbiota Transplantation (FMT)—where healthy bacteria are moved from one person to another to cure disease.

Currently, FMT relies on finding a random healthy donor, which can be risky and unpredictable. With RAcolony, scientists could potentially build a "personal backup" of a person's own healthy gut bacteria. If that person gets sick later, they could use their own pre-saved, perfectly identified bacteria to fix their gut, avoiding the risks of using someone else's microbes. The paper concludes that while there is still work to be done (like figuring out how to handle bacteria that stick together in clumps), RAcolony is a powerful new tool that makes the invisible world of our gut much easier to see, understand, and use.

Technical Summary: RAcolony – A Multimodal Platform for Personalized Gut Microbiome Biobanks

Problem Statement
The construction of personalized Gut Microbiome Biobanks (GMBs) is essential for advancing host-microbiome interaction research and developing precise therapeutic interventions, such as Fecal Microbiota Transplantation (FMT) and Synthetic Microbial Communities (SynComs). However, current isolation workflows are hindered by labor-intensive, low-throughput manual processes. Conventional culturomics often rely on phenotypic screening and repeated streaking, leading to the repetitive picking of highly redundant colonies (same species or strains) while overlooking slow-growing or fastidious bacteria. Furthermore, existing automated or single-cell sorting methods often lack chemotaxonomic fingerprinting, resulting in high redundancy and the failure to capture low-abundance taxa. There is an urgent need for an automated, high-precision technology that can efficiently discriminate bacterial strains and dereplicate colonies to maximize the recovery of microbial diversity.

Methodology
The authors developed RAcolony, an intelligent, high-throughput colony screening platform that integrates automated microscopic imaging with Raman spectroscopy within an anaerobic workstation. The system operates through a sequential workflow:

Localization and Imaging: Colonies on Petri dishes are localized and coded. High-resolution images are captured using macroscopic and microscopic modules (up to 10 μm resolution).
Spectral Acquisition: Raman spectra are collected for each colony using a 785 nm laser, providing label-free, non-destructive chemotaxonomic fingerprints based on biomacromolecules (proteins, lipids, nucleic acids).
Data Processing and Dereplication:
- Image Analysis: A deep learning-based segmentation algorithm (utilizing a pre-trained CNN and GPU acceleration) extracts 22 discriminatory morphological features (e.g., circularity, area, grayscale).
- Spectral Analysis: Raman spectra undergo baseline correction and normalization. A greedy iterative algorithm based on sample-space min–max distance is employed to optimize data point selection, mitigating signal-to-noise ratio (SNR) interference.
- Integration: Deep manifold learning (t-SNE) maps image and spectral features into a low-dimensional space. The system estimates the degree of differentiation ( $N$ ) to determine the number of distinct categories ( $2N$ ) for dereplication.
Automated Picking: Selected colonies are automatically picked and transferred to 96-well plates for cultivation.
Validation: The platform was validated using defined microbial communities (18 strains) and fecal samples from healthy donors. Comparisons were made against manual picking and random selection strategies. Taxonomic assignment was performed via 16S rRNA gene sequencing (Sanger for isolates, Illumina/Nanopore for community analysis), with species defined at <98.65% similarity and strains at 100% similarity.

Key Contributions

Multimodal Integration: RAcolony uniquely combines morphological imaging with Raman spectroscopic fingerprinting, overcoming the limitations of morphology-only identification where phylogenetically distinct species may appear similar.
Strain-Level Resolution: The platform demonstrates the ability to discriminate bacteria at the strain level by detecting subtle variations in Raman spectral peaks (e.g., secondary metabolite accumulation) and morphological heterogeneity, even among colonies of the same species.
Algorithmic Dereplication: The development of a specific algorithm that prioritizes diversity in colony selection, effectively reducing redundancy while ensuring balanced representation across taxa.
Scalable Automation: The system operates at a throughput of 1,000 colonies per hour within a controlled anaerobic environment, significantly reducing labor and time compared to manual methods.

Results

Platform Optimization: The MS medium was identified as optimal for both bacterial growth and Raman signal quality (highest SNR). A 3-day cultivation period was established as the standard, during which colony morphology and spectra remained stable.
Efficiency in Defined Communities: In experiments with a defined 18-strain community, RAcolony achieved species coverage rates of 76.5%–87.5% with redundancy rates of 100%–150%. In contrast, manual and random selection methods yielded significantly lower coverage (29.4%–75%) and much higher redundancy (210%–600%).
Personalized Biobank Construction:
- Using RAcolony on fecal samples from 7 donors, the platform screened 4,527 colonies, selecting 400 isolates. This yielded 99 distinct bacterial species (including 17 potential novel taxa) with an average redundancy rate of 160% ± 90%.
- The isolates captured 90.6% of the taxa present on the culture plates and 45.8% of species-level diversity from the original fecal samples.
- At the strain level, the platform identified 195 distinct strains from 501 isolates across 4 different media types, covering 80.9% of species and 55.1% of Amplicon Sequence Variants (ASVs) found in the original fecal samples.
Novelty: The study resulted in the isolation of 167 bacterial species in total (across 8 personalized GMBs), including 27 potentially novel taxa, 9 of which were characterized and proposed with new names.

Significance and Claims
The paper posits that RAcolony addresses critical bottlenecks in the construction of personalized GMBs by significantly reducing sampling redundancy and labor intensity while enhancing the recovery of gut microbial diversity.

Advancement over Conventional Methods: The authors claim RAcolony represents a significant advancement over traditional manual methods (which often exhibit redundancy rates exceeding 1000%) and previous automated platforms that rely solely on single-modal identification.
Clinical Relevance: By enabling the efficient, non-destructive isolation of an individual's own gut microbial strains, RAcolony facilitates the creation of autologous FMT banks and precise SynComs. This offers a strategic alternative to donor-dependent FMT, potentially circumventing donor scarcity, safety risks (e.g., multidrug-resistant organism transmission), and standardization issues.
Strain-Specific Utility: The platform's ability to distinguish strains is highlighted as crucial for probiotic screening and therapeutic applications, as functionality is often strain-dependent.
Scalability: The authors conclude that RAcolony is a validated, scalable platform that paves the way for the acquisition of high-quality, origin-traceable gut bacterial resources, thereby propelling the advancement of precision microbiome-based therapeutics.