Analyzing protein dynamics from fluorescence intensity traces using unsupervised deep learning network

Dual-Label Single-Molecule Imaging Method for Quantifying Apparent Fluorescence Efficiency of Fluorescent Proteins

Fluorescent proteins (FPs) are indispensable tools for life science research, crucial for quantitative analyses, such as protein stoichiometry determination, signal transduction, and protein-protein interactions. In this work, we introduce a new method of dual-color single-molecule imaging to assess the apparent fluorescence efficiency of FPs (DC-FEFP). By integrating high signal-to-noise ratio (SNR) FPs or self-labeling tags, this approach enables us to precisely quantify the fluorescence efficiency of various FPs at the single-molecule level in both living and fixed cells. With DC-FEFP, we found that mNeonGreen has the highest fluorescence efficiency among three commonly used FPs in living cells, as well as the significant impact of fixation on the photophysical properties of FPs. DC-FEFP is a high-precision and versatile tool for quantifying the fluorescence efficiency of FPs. It is capable of providing accurate information to calibrate protein stoichiometry based on single-molecule imaging.

Holistic Prediction of AuNP Aggregation in Diverse Aqueous Suspensions Based on Machine Vision and Dark-Field Scattering Imaging

The localized surface-plasmon resonance of the AuNP in aqueous media is extremely sensitive to environmental changes. By measuring the signal of plasmon scattering light, the dark-field microscopic (DFM) imaging technique has been used to monitor the aggregation of AuNPs, which has attracted great attention because of its simplicity, low cost, high sensitivity, and universal applicability. However, it is still challenging to interpret DFM images of AuNP aggregation due to the heterogeneous characteristics of the isolated and discontinuous color distribution. Herein, we introduce machine vision algorithms for the training of DFM images of AuNPs in different saline aqueous media. A visual deep learning framework based on AlexNet is constructed for studying the aggregation patterns of AuNPs in aqueous suspensions, which allows for rapid and accurate identification of the aggregation extent of AuNPs, with a prediction accuracy higher than 0.96. With the aid of machine learning analysis, we further demonstrate the prediction ability of various aggregation phenomena induced by both cation species and the concentration of the external saline solution. Our results suggest the great potential of machine vision frameworks in the accurate recognition of subtle pattern changes in DFM images, which can help researchers build predictive analytics based on DFM imaging data.

Feature Selection Assists BLSTM for the Ultrasensitive Detection of Bioflavonoids in Different Biological Matrices Based on the 3D Fluorescence Spectra of Gold Nanoclusters

Rapid and on-site qualitative and quantitative analysis of small molecules (including bioflavonoids) in biofluids are of great importance in biomedical applications. Herein, we have developed two deep learning models based on the 3D fluorescence spectra of gold nanoclusters as a single probe for rapid qualitative and quantitative analysis of eight bioflavonoids in serum. The results proved the efficiency and stability of the random forest-bidirectional long short-term memory (RF-BLSTM) model, which was used only with the most important features after deleting the unimportant features that might hinder the performance of the model in identifying the selected bioflavonoids in serum at very low concentrations. The optimized model achieves excellent overall accuracy (98-100%) in the qualitative analysis of the selected bioflavonoids. Next, the optimized model was transferred to quantify the selected bioflavonoids in serum at nanoscale concentrations. The transferred model achieved excellent accuracy, and the overall determination coefficient (R²) value range was 99-100%. Furthermore, the optimized model achieved excellent accuracies in other applications, including multiplex detection in serum and model applicability in urine. Also, LOD in serum at nanoscale concentration was considered. Therefore, this approach opens the window for qualitative and quantitative analysis of small molecules in biofluids at nanoscale concentrations, which may help in the rapid inclusion of sensor arrays in biomedical and other applications.

Deep learning in single-molecule imaging and analysis: recent advances and prospects

Single-molecule microscopy is advantageous in characterizing heterogeneous dynamics at the molecular level. However, there are several challenges that currently hinder the wide application of single molecule imaging in bio-chemical studies, including how to perform single-molecule measurements efficiently with minimal run-to-run variations, how to analyze weak single-molecule signals efficiently and accurately without the influence of human bias, and how to extract complete information about dynamics of interest from single-molecule data. As a new class of computer algorithms that simulate the human brain to extract data features, deep learning networks excel in task parallelism and model generalization, and are well-suited for handling nonlinear functions and extracting weak features, which provide a promising approach for single-molecule experiment automation and data processing. In this perspective, we will highlight recent advances in the application of deep learning to single-molecule studies, discuss how deep learning has been used to address the challenges in the field as well as the pitfalls of existing applications, and outline the directions for future development.

Toward high-throughput oligomer detection and classification for early-stage aggregation of amyloidogenic protein

Aggregation kinetics of proteins and peptides have been studied extensively due to their significance in many human diseases, including neurodegenerative disorders, and the roles they play in some key physiological processes. However, most of these studies have been performed as bulk measurements using Thioflavin T or other fluorescence turn-on reagents as indicators of fibrillization. Such techniques are highly successful in making inferences about the nucleation and growth mechanism of fibrils, yet cannot directly measure assembly reactions at low protein concentrations which is the case for amyloid-β (Aβ) peptide under physiological conditions. In particular, the evolution from monomer to low-order oligomer in early stages of aggregation cannot be detected. Single-molecule methods allow direct access to such fundamental information. We developed a high-throughput protocol for single-molecule photobleaching experiments using an automated fluorescence microscope. Stepwise photobleaching analysis of the time profiles of individual foci allowed us to determine stoichiometry of protein oligomers and probe protein aggregation kinetics. Furthermore, we investigated the potential application of supervised machine learning with support vector machines (SVMs) as well as multilayer perceptron (MLP) artificial neural networks to classify bleaching traces into stoichiometric categories based on an ensemble of measurable quantities derivable from individual traces. Both SVM and MLP models achieved a comparable accuracy of more than 80% against simulated traces up to 19-mer, although MLP offered considerable speed advantages, thus making it suitable for application to high-throughput experimental data. We used our high-throughput method to study the aggregation of Aβ40 in the presence of metal ions and the aggregation of α-synuclein in the presence of gold nanoparticles.

Deep Learning-Based Sensor Array: 3D Fluorescence Spectra of Gold Nanoclusters for Qualitative and Quantitative Analysis of Vitamin B6 Derivatives

Vitamin B₆ derivatives (VB6Ds) are of great importance for all living organisms to complete their physiological processes. However, their excess in the body can cause serious problems. What is more, the qualitative and quantitative analysis of different VB6Ds may present significant challenges due to the high similarity of their chemical structures. Also, the transfer of deep learning model from one task to a similar task needs to be present more in the fluorescence-based biosensor. Therefore, to address these problems, two deep learning models based on the intrinsic fingerprint of 3D fluorescence spectra have been developed to identify five VB6Ds. The accuracy ranges of a deep neural network (DNN) and a convolutional neural network (CNN) were 94.44-97.77% and 97.77-100%, respectively. After that, the developed models were transferred for quantitative analysis of the selected VB6Ds at a broad concentration range (1-100 μM). The determination coefficient (R²) values of the test set for DNN and CNN were 93.28 and 97.01%, respectively, which also represents the outperformance of CNN over DNN. Therefore, our approach opens new avenues for qualitative and quantitative sensing of small molecules, which will enrich fields related to deep learning, analytical chemistry, and especially sensor array chemistry.

An antibody-free liver cancer screening approach based on nanoplasmonics biosensing chips via spectrum-based deep learning

The clinical needs of rapidly screening liver cancer in large populations have asked for a facile and low-cost point-of-care testing (POCT) method. We present a nanoplasmonics biosensing chip (NBC) that would empower antibody-free detection with simplified analysis procedures for POCT. The cheaply fabricable NBC consists of multiple silver nanoparticle-decorated ZnO nanorods on cellulose filter paper and would enable one-drop blood tests through surface-enhanced Raman spectroscopy (SERS) detection. In this work, utilizing such an NBC and deep neural network (DNN) modeling, a direct serological detection platform was constructed for automatically identifying liver cancer within minutes. This chip could enhance Raman signals enough to be applied to POCT. A classification DNN model was established by spectrum-based deep learning with 1140 serum SERS spectra in equal proportions from hepatocellular carcinoma (HCC) patients and healthy individuals, achieving an identification accuracy of 91% on an external validation set of 100 spectra (50 HCC versus 50 healthy). The intelligent platform, based on the biosensing chip and DNN, has the potential for clinical applications and generalizable use in quickly screening or detecting other types of cancer.