Machine learning expands single-molecule analysis accuracy and accessibility

Adapted from Center for RNA Biomedicine Press Release, November 19, 2020

The observation of single biomolecules in real-time is crucial for our understanding of the cellular biology that is assembled from these molecules, be they DNA, RNA or protein. The recent development of an array of tools and techniques for single-molecule analysis allows studies at an extremely small scale (nanometers, or 10-9 meters) over short periods of time (from a few milliseconds to a second).

Illustration: Deep learning assisted single molecule fluorescence microscopy data analysis

However, until now, most of such observations required tedious and time-consuming manual data processing of thousands of single molecules. A team of University of Michigan (U-M) Department of Chemistry and Department of Physics scientists, spearheaded by graduate students Jieming Li, now Ph.D. in Chemistry, and Leyou Zhang, now Ph.D. in Physics, developed a deep learning algorithm to analyze data emerging from a single-molecule microscope. The results from this collaboration are published in Nature Communications (November 2020).

The key to their automatic data analysis workflow for single molecules (called “AutoSiM”) is to apply a deep learning algorithm, representing a special class of machine learning or artificial intelligence that contains many layers of a neural network, to single molecule fluorescence microscopy data. First, the pattern of the fluorescence signal is recognized to classify the molecule as either relevant or not. Then, the algorithm distinguishes which segment within the fluorescence signal should be included for further data interpretation. For training, the scientists fed the program with a large dataset that was already analyzed by human experts to teach the algorithm to recognize the correct pattern of fluorescence signals, which reflect the behavior of bio-molecules. Once trained on one dataset, it can quickly be adapted to any new dataset by a short “transfer learning”.

There are three main advantages to using AutoSiM: it is faster, significantly cuts down on research time, and reduces costs. “One of my goals is to free up researchers from doing tedious data analysis so they can better focus on their exciting scientific inquiries,” says Li. With the network, data analysis is more consistent than possible across human researchers (due to individual biases), and is free from data entry errors. The concordance between AutoSiM and manual selection is about 90% and is completely reproducible from day to day.

AutoSiM was developed on two datasets from the lab of Nils Walter, U-M Francis S. Collins Collegiate Professor of Chemistry, Biophysics, and Biological Chemistry. One dataset comprised kinetic fingerprints of single molecules termed SiMREPS time traces. The SiMREPS assay distinguishes surface-immobilized mutant DNA biomarkers of disease from wild-type on the basis of distinguishable fluorescence kinetics when a probe interacts with the mutant sequence, wild-type sequence, or surface itself. This molecular diagnostics tool is now being commercialized by aLight Sciences Corp., co-founded by Walter and co-author Alexander Johnson Buck. The other set consisted of time traces characterizing the shape changes of 4 different biomolecular complexes using single molecule FRET (smFRET). Since its development over two decades ago, smFRET has been used to study the dynamics of many biomolecules at the nanometer scale, especially conformational changes involving nucleic acids and/or proteins.

The workflow allows for transfer learning, which means that it can be adapted to new systems by learning from small additional datasets, expanding the original capabilities of the algorithm. “This is like our human brain that learns a lot of new information at a young age, but can also readily adopt new input over its entire lifespan,” says Walter. “It’d be great if others would include their own data to train and use AutoSiM, so the capabilities of the network can be expanded,” explains Li.

The newly released software is available and free for academic purposes, through the U-M Library Deep Blue data repository.

The University of Michigan (U-M) supports state-of-the-art single molecular microscopy at the Single Molecule Analysis in Real-Time (SMART) Center, one of the two core facilities of the U-M Center for RNA Biomedicine. The SMART Center is a shared-use facility providing university researchers with single molecule detection and manipulation tools to track and analyze biomolecules with unprecedented detail. The SMART Center provides access to instrumentation, including single molecule spectroscopy and imaging, laser tweezers, and atomic force microscopy; as well as experienced support in experimental planning and analysis.

Paper cited:
Classification and segmentation of single-molecule fluorescence time traces with deep learning, Jieming Li, Leyou Zhang, Alexander Johnson-Buck and Nils G. Walter, Nature Communications, (2020)11:5833,

Methods Editorial: “Introduction to “Convergence of Science and Technology: Fluorescent Resolution of Single RNA Molecules”

As Phillip Sharp, Tyler Jacks and Susan Hockfield of MIT posit in a 2016 report [1]: “The life sciences are in the midst of a revolution… The Convergence Revolution promises to enhance quality of life worldwide … as a result of the sharing of methods and ideas by chemists, physicists, computer scientists, engineers, mathematicians, and life scientists across multiple fields and industries. It is the integration of insights and approaches from historically distinct scientific and technological disciplines.” Perhaps nowhere is this both more apparent and more needed than in taking advantage of the rapid expansion of the RNA biosciences universe and the resulting opportunities for translating discoveries into precision medicines [2]. Converging technological advances, ranging from single-cell RNA sequencing to single-molecule particle tracking at super-resolution, are laying the foundation for reaching an entirely new level of insight into cellular physiology that eluded us just a few years ago [3]. Beyond the emerging catalogues of cellular contents, mechanistic studies of the diverse modes of action of RNAs and their interaction partners are handing us the keys for accelerating progress toward shaping our own destiny. The better we understand the scope and limitations of these new tools, the better we will be equipped to wield them effectively.


Read full editorial in the journal Methods : Convergence of science and technology: fluorescent resolution of single RNA molecules

PDF version

A huge milestone for RNA research at Michigan! U-M Biosciences Initiative invests $45M in ‘groundbreaking’ research (Michigan News)

October 29, 2018

University of Michigan Biosciences Initiative

ANN ARBOR—A new center for the study of concussions, an institute for global change biology, and a facility to advance the new field of cryo-electron tomography are among the University of Michigan projects to be funded in the first round of investments from President Mark Schlissel’s Biosciences Initiative.

Read full article in Michigan News here : U-M Biosciences Initiative invests $45M in ‘groundbreaking’ research

Nature Reviews Research Highlight: “Signals from single molecules”

Mutated DNA strands — even those with only a change in a single nucleobase — can be the cause of serious health problems. Quantifying these mutant DNA strands is particularly challenging when they are present in low concentrations in a patient’s urine or blood samples and must be detected against a large background concentration of wild-type DNA. Now, writing in the Journal of the American Chemical Society, Nils Walter and co-workers report the development of an amplification-free method that they call single-molecule recognition through equilibrium Poisson sampling (SiMREPS).

“We were seeking a powerful tool to detect mutations that are cancer causing, or that make cancer harder to treat,” explains Walter. “One such mutation is found in the epidermal growth factor receptor (EGFR), where a cytosine-to-thymine mutation in the DNA results in substitution of a threonine with a methionine in the protein. The ultimate result of this change is that certain classes of tyrosine kinase inhibitor (TKI) drugs no longer work.”

Read full research highlight in the journal Nature Reviews Materials : Signals from single molecules

PDF version

Nils quoted in Genetic Engineering & Biotechnology news article “Get a Solid Grasp on Tissue RNA Analysis

“Modern techniques like RNA sequencing run into the challenge that low-expressed RNAs are difficult to detect, but often they are more important than the abundant ones,” observes Dr. Walter. “We need tools to look at them.” In a recent study, Dr. Walter and colleagues developed fluorescence-based tools to examine the subcellular location, integrity, and activity of microRNA at single-molecule resolution. “We can localize single microRNA molecules in cells, track them, measure their diffusion constants, and ask how quickly they move between sub-compartments,” asserts Dr. Walter

Full text in pdf format.

Nature Cell Biology “News and Views” : RNA takes over control of DNA break repair

Francesca Storici and Ailone E. Tichon

Small RNAs generated at DNA break sites are implicated in mammalian DNA repair. Now, a study shows that following the formation of DNA double-strand breaks, bidirectional transcription events adjacent to the break generate small RNAs that trigger the DNA damage response by local RNA:RNA interactions.

Maintenance of DNA integrity is crucial for a cell to have a healthy life and for transmission of accurate genetic information to its progeny. Exogenous agents, including radiation or chemicals, as well as endogenous sources, such as reactive oxygen species or defects in DNA metabolism, pose threats to genome stability. Among the most dangerous DNA lesions are DNA double-strand breaks (DSBs), which if not properly and timely sealed can become sites of mutations or chromosomal rearrangements — well-known hallmarks of cancer and other genetic disorders1. The process of DSB repair is one of the most extensively studied mechanisms of DNA repair, yet much remains to be understood about its players and dynamics.
The DNA damage response (DDR) is a complex network of cellular pathways that detect DNA lesions and organize a response signal cascade to repair the DNA. In this issue of Nature Cell Biology, Michelini et al.2 uncover an aspect of the DDR in which RNAs are the directors. The study shows that sequencespecific RNA:RNA interactions orchestrate the DDR in response to DSBs.

Full text in pdf format.

News and Views pice in reference to the Walter Lab NCB paper (ref. 164 at

The drawing shows three different DNA DSB sites in a cell nucleus, depicted as interrupted, thick, parallel strands in blue, green and orange, respectively. The work by Michelini et al.2 suggests that every different DNA DSB site has its unique set of RNA:RNA interactions (shown as small, sandwiched, purple or red rectangles at the two DSB sites on the left) between its specific dilncRNAs and small non-coding RNAs termed DDRNAs (both shown as light blue, green or yellow lines), which originate from dilncRNAs at each DSB site. Such local RNA:RNA interactions are signals to activate the DDR by recruitment of the early DDR factor 53BP1 (transparent ovals). The dilncRNAs generated from DSB ends (thicker lines) were found to be more abundant than those generated towards DSB ends (thinner lines). ASOs can be designed to specifically block DSB repair at a chosen locus, without interfering with DSB repair at other loci within the same nucleus. Here, ASOs (shown as short, thick, brown lines) are specific to dilncRNA sequences of the DSB site on the right and block formation and function of DDRNAs at this DSB site by interacting (small, sandwiched, grey rectangles) with the complementary dilncRNAs.

Nature Reviews Research Highlight: DNA NANOTECHNOLOGY | Head over heels

Dynamic DNA nanotechnology enables the design of DNA-based nanomachines, such as molecular motors or nanorobots. However, most DNA nanomachines operate on a slow timescale, ranging from minutes to hours. Now, Nils Walter and colleagues, writing in Nature Nanotechnology, have made a single-stranded DNA walker that moves by performing cartwheels and at considerably faster speeds than previously reported DNA walkers.

Credit: Rachael Tremlett/Macmillan Publishers Limited

Read full research highlight in the journal Nature Reviews Materials : Head over heels

PDF version


Study finds snap-lock mechanism in bacterial riboswitch

Using single-molecule imaging, scientists from Rice University and the University of Michigan determined that the T-box riboswitch that regulates production of glycine in Bacillus subtilis becomes locked into the “on” position via a snap-lock mechanism. The lock is engaged when an L-shaped arm of uncharged tRNA snaps into position. When the arm is “charged” with a glycine molecule, the arm cannot lock into position, and the switch remains “off.” (Image courtesy of E. Nikonowicz/Rice University)

Finding by Rice, Michigan chemists could point way to new antibiotics
HOUSTON — (May 21, 2018) — In a discovery that points to potential new antibiotic medicines, scientists from Rice University and the University of Michigan have deciphered the workings of a common but little-understood bacterial switch that cuts off protein production before it begins.

Many gram-positive bacteria use T-box riboswitches to regulate production of proteins that utilize amino acids, the basic building blocks of all proteins. A study in Nature Communications describes how one of these switches, a glycine regulator in Bacillus subtilis, flips and locks into the “on” position via a snap-lock mechanism. Engaging the lock increases production of proteins that utilize glycine, the simplest amino acid. Researchers also detailed the switch’s “off” position: A single glycine at the tip of the locking arm blocks protein production.


Read full text here: Study finds snap-lock mechanism in bacterial riboswitch

Nature Research Highlight: Gymnastic feats help DNA ‘walker’ set speed record

A molecular motor flips end over end to cover 300 nanometres a minute.

A short segment of DNA anchors one end of itself and then flings the other end forward. Credit: Zhuoru Li

An acrobatic DNA segment that cartwheels across a surface is one of the fastest ‘walking’ molecules yet designed.

Researchers seeking to design nanorobots have synthesized DNA molecules that move autonomously. But most of these ‘walkers’ require several minutes to take a single step.

Nils Walter at the University of Michigan in Ann Arbor and his colleagues built a DNA walker that moves along a grid of ‘footholds’, which are made of complementary DNA building blocks called nucleotides. Each end of the walker has sequences of nucleotides that it uses to attach itself to the grid. Then it flips end over end to attach itself to another foothold.

The molecule can flip as many as 43 times per minute, covering a distance of 300 nanometres. This is an order of magnitude faster than the pace of other types of DNA walker.

Read full research highlight in the journal Nature : Gymnastic feats help DNA ‘walker’ set speed record

Tiny nanomachine successfully completes test drive

Researchers build a one-wheeled vehicle out of DNA rings

The two rings are linked like a chain and can well be recognized. At the centre there is the T7 RNA Polymerase.
Credit: Copyright Julián Valero

Scientists have used nanostructures to construct a tiny machine that constitutes a rotatory motor and can move in a specific direction. The researchers used circular structures from DNA.

Together with colleagues from the USA, scientists from the University of Bonn and the research institute Caesar in Bonn have used nanostructures to construct a tiny machine that constitutes a rotatory motor and can move in a specific direction. The researchers used circular structures from DNA. The results will now be presented in the journal Nature Nanotechnology.

Read full text here: Tiny nanomachine successfully completes test drive

Built for speed: DNA nanomachines take a (rapid) step forward

When it comes to matching simplicity with staggering creative potential, DNA may hold the prize. Built from an alphabet of just four nucleic acids, DNA provides the floorplan from which all earthly life is constructed.

Through a process known as strand displacement, a tiny walking device composed of DNA moves across a surface in a cartwheeling motion. The new device performed this feat more rapidly than any DNA walker designed to date. Credit: Nature Nanotechnology/Nils Walter

But DNA’s remarkable versatility doesn’t end there. Researchers have managed to coax segments of DNA into performing a host of useful tricks. DNA sequences can form logical circuits for nanoelectronic applications. They have been used to perform sophisticated mathematical computations, like finding the optimal path between multiple cities. And DNA is the basis for a new breed of tiny robots and nanomachines. Measuring thousands of times smaller than a bacterium, such devices can carry out a multitude of tasks.

In new research, Hao Yan of Arizona State University and his colleagues describe an innovative DNA walker, capable of rapidly traversing a prepared track. Rather than slow, tentative steps across a surface, the DNA acrobat cartwheels head over heels, covering ground 10- to 100-fold faster than previous devices.

“It is exciting to see that DNA walkers can increase their speed significantly by optimizing DNA strand length and sequences, the collaborative effort really made this happen,” Yan said.

Yan is the Milton D. Glick Distinguished Professor of Chemistry and Biochemistry at ASU and director of the Biodesign Center for Molecular Design and Biomimetics.

The study was led by Nils G. Walter, Francis S. Collins Collegiate Professor of Chemistry, Biophysics & Biological Chemistry, founding director of the Single Molecule Analysis in Real-Time (SMART) Center and founding co-director of the Center for RNA Biomedicine at the University of Michigan, and his team, along with collaborators from the Wyss Institute, the Dana Farber Cancer Institute and the Department of Biological Chemistry at Harvard (all in Boston, Massachusetts).

“The trick was to make the walker go head over heels, which is so much faster than the hopping used before—just as you would see in a kung fu action movie where the hero speeds up by cartwheeling to catch the villain,” says Walter.

The improvements in speed and locomotion displayed by the new walker should encourage further innovations in the field of DNA nanotechnology.

The group’s findings appear in the advanced online issue of the journal Nature Nanotechnology.

Read full text here: Building Motors to Drive Nanorobots