Mahyar Dahmardeh M.Sc.

Self-supervised machine learning pushes the sensitivity limit in label-free detection of single proteins below 10 kDa

Mahyar Dahmardeh, Houman Mirzaalian Dastjerdi, Hisham Mazal, Harald Köstler, Vahid Sandoghdar

Interferometric scattering (iSCAT) microscopy is a label-free optical method capable of detecting single proteins, localizing<br>their binding positions with nanometer precision, and measuring their mass. In the ideal case, iSCAT is limited by shot noise<br>so that collection of more photons should allow its detection sensitivity to biomolecules of arbitrarily low mass. However, a<br>number of technical noise sources combined with speckle-like background fluctuations have restricted the detection limit in<br>iSCAT. Here, we show that an unsupervised machine learning isolation forest algorithm for anomaly detection pushes the<br>mass sensitivity limit by a factor of four to below 10 kDa. We implement this scheme both with a user-defined feature matrix<br>and a self-supervised FastDVDNet and validate our results with correlative fluorescence images recorded in total internal<br>reflection mode. Our work opens the door to the optical detection of small traces of disease markers such as alpha-synuclein,<br>chemokines, and cytokines.

Optimized analysis for sensitive detection and analysis of single proteins via interferometric scattering microscopy

Houman Mirzaalian Dastjerdi, Mahyar Dahmardeh, André Gemeinhardt, Reza Gholami Mahmoodabadi, Harald Köstler, Vahid Sandoghdar

It has been shown that interferometric detection of Rayleigh scattering (iSCAT) can reach an exquisite sensitivity for label-free detection of nano-matter, down to single proteins. The sensitivity of iSCAT detection is intrinsically limited by shot noise, which can be indefinitely improved by employing higher illumination power or longer integration times. In practice, however, a large speckle-like background and technical issues in the experimental setup limit the attainable signal-to-noise ratio. Strategies and algorithms in data analysis are, thus, crucial for extracting quantitative results from weak signals, e.g. regarding the mass (size) of the detected nano-objects or their positions. In this article, we elaborate on some algorithms for processing iSCAT data and identify some key technical as well as conceptual issues that have to be considered when recording and interpreting the data. The discussed methods and analyses are made available in the extensive python-based platform, PiSCAT.