23 December 2025 in ChemRxiv
Terminally labeled DNA oligonucleotides have wide applications in modern biology and biotechnological applications. It has been observed that the fluorescent intensity of light released from these fluorescent labels is heavily influenced by the terminal sequence of nucleotides. Recent studies have assayed and published the raw fluorescent values of Cy3 and Cy5 as a function of the most adjacent 5 nucleotides resulting in 1024 data points. While experimentally tractable, an increase in the sequence space will vastly increase the experimental and time cost. Machine Learning is well suited to addressing the issue of experimental tractability however there is a wide design space in the choice of algorithms. In this work we use classic machine learning models such as Support Vector Machine, Multilayer Perceptrons and Random Forests to both predict the raw intensity value and classify the intensity magnitude of the fluorophore using the sequence as input. We demonstrate that the performance of these models is heavily dependent on the numerical transformation of the sequence and that Random Forest consistently outperforms all other models in both regression and classification tasks irrespective of the sequence transformation.
22 December 2025 in bioRxiv
Background: Super Resolution Microscopy (SRM) is a powerful method in quantitative bioscience that allows interrogation of nanoscale details. These methods require extensive imaging times on the microscope resulting in data sets on the scale of gigabytes. In order to reduce imaging times, the concentration of emitters can be increased, however that results in overlapping emitters rendering the isolation of single emitters extremely difficult. Statistical methods have been developed to deconvolute overlapping emitters, however they require parameter optimization and user expertise. Recently, Machine Learning (ML) has been developed to automate this analysis but often require larger Convolutional Neural Networks (CNN). While powerful, such models require compute and storage that would make pushing these models to compute limited devices difficult. To address this, we investigate if the dense multi-emitter localization capacity of a larger model, Deep Residual Stochastic Optical Reconstruction Microscopy (DRL-STORM), can be transferred to a smaller model, Super Resolution Convolutional Neural Network (SRCNN). Results: Knowledge transfer from DRL-STORM to SRCNN did not result in an improvement of multi-emitter localization performance of SRCNN. Hint Learning (HL) was performed to facilitate knowledge transfer in a more deliberate manner. SRCNN demonstrated a limited capacity to learn an intermediate representation of the input image in the same manner as DRL-STORM and resultantly did not perform any better in its task. Conclusions: Knowledge transfer was not successful between DRL-STORM and SRCNN, but evidence suggests that it is possible and may require another model besides SRCNN. A future work will investigate if hyper-parameter optimization results in greater knowledge distillation between DRL-STORM and SRCNN. SRCNN may not be ideal for multi-emitter localizations, but it can still prove effective for SRM data analysis at emitter concentrations typical of SRM experiments and is uniquely suited as a neural network model that can be deployed in compute limited settings.