Label-free identification of protein aggregates using deep learning

Protein misfolding and aggregation play central roles in the pathogenesis of various neurodegenerative diseases (NDDs), including Huntington’s disease, which is caused by a genetic mutation in exon 1 of the Huntingtin protein (Httex1). The fluorescent labels commonly used to visualize and monitor the dynamics of protein expression have been shown to alter the biophysical properties of proteins and the final ultrastructure, composition, and toxic properties of the formed aggregates. To overcome this limitation, we present a method for label-free identification of NDD-associated aggregates (LINA). Our approach utilizes deep learning to detect unlabeled and unaltered Httex1 aggregates in living cells from transmitted-light images, without the need for fluorescent labeling. Our models are robust across imaging conditions and on aggregates formed by different constructs of Httex1. LINA enables the dynamic identification of label-free aggregates and measurement of their dry mass and area changes during their growth process, offering high speed, specificity, and simplicity to analyze protein aggregation dynamics and obtain high-fidelity information.

For all statistical analyses, confirm that the following items are present in in the figure legend, table legend, main text, or or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as as a discrete number and unit of of measurement A statement on on whether measurements were taken from distinct samples or or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of of all covariates tested
A description of of any assumptions or or corrections, such as as tests of of normality and adjustment for multiple comparisons A full description of of the statistical parameters including central tendency (e.g.means) or or other basic estimates (e.g.regression coefficient) AND variation (e.g. standard deviation) or or associated estimates of of uncertainty (e.g.confidence intervals) For null hypothesis testing, the test statistic (e.g.F, t, r) with confidence intervals, effect sizes, degrees of of freedom and P value noted Give P values as exact values whenever suitable.
For Bayesian analysis, information on on the choice of of priors and Markov chain Monte Carlo settings For hierarchical and complex designs, identification of of the appropriate level for tests and full reporting of of outcomes Estimates of of effect sizes (e.g.Cohen's d, Pearson's r), ), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above.

Software and code
Policy information about availability of of computer code

Data analysis
For manuscripts utilizing custom algorithms or or software that are central to to the research but not yet described in in published literature, software must be be made available to to editors and reviewers.We We strongly encourage code deposition in in a community repository (e.g.GitHub).See the Nature Portfolio guidelines for submitting code & software for further information.
Prof. Kristin Grußmayer, Prof. Hilal Lashuel, Prof. Aleksandra Radenovic Oct 23, 2023 Custom LabVIEW code (2021 version) was used for data acquisition on on our multi-modal, multi-plane microscope.The automatic scanning mode and the microscope are described in in the manuscript.
Data analysis was performed using custom MATLAB (R2021a), available here: https://c4science.ch/source/TomPhaseRet/, to to retrieve the phase information from the brightfield images and produce quantitative phase images.We We used Fiji (v2.9.0) scripts to to produce maximum z-zprojection fluorescence images, to to segment these images using Otsu thresholding and produce the labels for pixel classification, to to prepare the color-coded maximum z-projection phase image shown in in Fig. 2a, and for the image processing in in Fig. 2b.Fiji was also used to to segment network predictions and produce masks which are used to to measure the area, circularity or or dry mass of of aggregates.Python (v3.7) was used for data analysis and plotting.Python was also used for neural network training, along with Tensorflow 2.8, Keras 2.8, CUDA 11.1 and CUDNN 8. HEK293 cells were identified by by the vendor: https://www.atcc.org/products/crl-11268(karyotyping) and were not authenticated for this study in in particular.
The HEK293 cells were tested negative to to mycoplasma contamination.HeLa cells were not tested for mycoplasma after purchase from the vendor.
No No commonly misidentified cell lines were used in in the study.