Carnegie Mellon University
February 06, 2024

New Methods Allow for Improved Molecular Monitoring

By Kirsten Heuring

Heidi Opdyke
  • Interim Director of Communications, MCS

Tracking a single microscopic particle in space and time creates computational challenges. An international team of biological physics researchers has developed a machine learning method to track multiple particles, including proteins binding transiently to DNA, simultaneously. Temporal analysis of relative distances (TARDIS) opens new possibilities for biological research including monitoring repairs at the cellular level.

"The limits with single particle tracking were always that you had to look at one particle at the same time, and with TARDIS, that's simply not a limit anymore," said Martens, a former postdoctoral fellow in Carnegie Mellon University's Department of Physics now at the University of Bonn. "Now what I can do is get a very precise view of how quick DNA repair is."

Single particle tracking has significant problems, Martens said. The method is not the most precise tool temporally or spatially and can create results that look plausible but are actually errors. Single particle tracking also does not tell researchers if multiple particles are crossing.

TARDIS allows for multiple particles within a liquid or within living cells to be tracked at the same time. The method can be done in a significantly shorter period of time with greater accuracy on particle location. When it fails or makes errors, the issues are more obvious.

Ulrike Endesfelder, professor of microbiology and biotechnology at the University of Bonn and former associate professor of physics at Carnegie Mellon University, said she has implemented TARDIS in combination with a previous molecule tracking software that she developed, swift, to ensure her lab gets the most precise and accurate data possible. Endesfelder's swift was developed as an advancement in single particle tracking. It can be used to track single molecules precisely, and it is commonly used in her lab.

"Together, TARDIS and swift are a perfect tracking software team — and my current best answer to the question of how to tackle complex single-molecule tracking scenarios," Endesfelder said. "TARDIS is so straightforward that it doesn't fail silently. It doesn't make any assumptions and thus is not biased."

"Designing tools like this is a benefit of working in an interdisciplinary team." — Koen Martens

Markus Deserno, professor of physics at Carnegie Mellon, assisted with some of the mathematics and probabilities behind TARDIS. He said he was excited to work with Martens and Endesfelder, and he looks forward to the potential of TARDIS to advance biological physics, the intersection where methods from physics are used to study biological systems on the molecular level to the scale of whole organisms.

"There is a huge and growing interest in recent years not just about where stuff is inside cells, but also how it's moving — meaning, we now want temporal information on top of the impressive spatial information," Deserno said. "Unfortunately, stitching different frames together to form a movie in which one can track what actor is moving where has historically been very difficult, due to numerous annoying artifacts and problems that unavoidably arise because observation times only happen at discrete intervals."

Deserno said the new TARDIS algorithm is impressively robust.

"As more people use it, we will learn a whole lot more about cellular dynamics, with the optical systems already in existence," he said.

Martens said he plans to use TARDIS to track how DNA repairs itself after damage. Because of TARDIS's versatility and user friendliness, he said he believes that the applications are broad, and it can be used in any field where researchers are tracking multiple particles in a liquid. The researchers have made TARDIS available on GitHub.

"I specifically made it so people who are interested in coding can do anything they want, but it's also a standalone program, so people who are not as experienced in computer science can just press a download button," Martens said. "A lot of researchers want to use really cool tools but don't know how, so designing tools like this is a benefit of working in an interdisciplinary team."

Martens and Endesfelder were joined by Bartosz Turkowyd and Johannes Hohlbein on "Temporal Analysis of Relative Distances (TARDIS) Is a Robust, Parameter-Free Alternative to Single-Particle Tracking," which was published in Nature Methods. The research was funded by Carnegie Mellon, the NSF AI Institute: Physics of the Future, the University of Bonn and the Alexander von Humboldt Foundation.

— Related Content —