Exploiting weak supervision to facilitate segmentation, classification, and analysis of microplastics (<100 µm) using Raman microspectroscopy images.

Reliable quantification and characterization of microplastics are necessary for large-scale and long-term monitoring of their behaviors and evolution in the environment. This is especially true in recent times because of the increase in the production and use of plastics during the pandemic. However, because of the myriad of microplastic morphologies, dynamic environmental forces, and time-consuming and expensive methods to characterize microplastics, it is challenging to understand microplastic transport in the environment. This paper describes a novel approach that compares unsupervised, weakly-supervised, and supervised approaches to facilitate segmentation, classification, and the analysis of <100 µm-sized microplastics without the use of pixel-wise human-labeled data. The secondary aim of this work is to provide insight into what can be accomplished when no human annotations are available, using the segmentation and classification tasks as use cases. In particular, the weakly-supervised segmentation performance surpasses the baseline performance set by the unsupervised approach. Consequently, feature extraction (derived from the segmentation results) provides objective parameters describing microplastic morphologies that will result in better standardization and comparisons of microplastic morphology across future studies. The weakly-supervised performance for microplastic morphology classification (e.g., fiber, spheroid, shard/fragment, irregular) also exceeds the performance of the supervised analogue. Moreover, in contrast to the supervised method, our weakly-supervised approach provides the benefit of pixel-wise detection of microplastic morphology. Pixel-wise detection is used further to improve shape classifications. We also demonstrate a proof-of-concept for distinguishing microplastic particles from non-microplastic particles using verification data from Raman microspectroscopy. As the automation of microplastic monitoring progresses, robust and scalable identification of microplastics based on their morphology may be achievable.

A Tale of Loops and Tails: The Role of Intrinsically Disordered Protein Regions in R-Loop Recognition and Phase Separation.

R-loops are non-canonical, three-stranded nucleic acid structures composed of a DNA:RNA hybrid, a displaced single-stranded (ss)DNA, and a trailing ssRNA overhang. R-loops perform critical biological functions under both normal and disease conditions. To elucidate their cellular functions, we need to understand the mechanisms underlying R-loop formation, recognition, signaling, and resolution. Previous high-throughput screens identified multiple proteins that bind R-loops, with many of these proteins containing folded nucleic acid processing and binding domains that prevent (e.g., topoisomerases), resolve (e.g., helicases, nucleases), or recognize (e.g., KH, RRMs) R-loops. However, a significant number of these R-loop interacting Enzyme and Reader proteins also contain long stretches of intrinsically disordered regions (IDRs). The precise molecular and structural mechanisms by which the folded domains and IDRs synergize to recognize and process R-loops or modulate R-loop-mediated signaling have not been fully explored. While studying one such modular R-loop Reader, the Fragile X Protein (FMRP), we unexpectedly discovered that the C-terminal IDR (C-IDR) of FMRP is the predominant R-loop binding site, with the three N-terminal KH domains recognizing the trailing ssRNA overhang. Interestingly, the C-IDR of FMRP has recently been shown to undergo spontaneous Liquid-Liquid Phase Separation (LLPS) assembly by itself or in complex with another non-canonical nucleic acid structure, RNA G-quadruplex. Furthermore, we have recently shown that FMRP can suppress persistent R-loops that form during transcription, a process that is also enhanced by LLPS via the assembly of membraneless transcription factories. These exciting findings prompted us to explore the role of IDRs in R-loop processing and signaling proteins through a comprehensive bioinformatics and computational biology study. Here, we evaluated IDR prevalence, sequence composition and LLPS propensity for the known R-loop interactome. We observed that, like FMRP, the majority of the R-loop interactome, especially Readers, contains long IDRs that are highly enriched in low complexity sequences with biased amino acid composition, suggesting that these IDRs could directly interact with R-loops, rather than being "mere flexible linkers" connecting the "functional folded enzyme or binding domains". Furthermore, our analysis shows that several proteins in the R-loop interactome are either predicted to or have been experimentally demonstrated to undergo LLPS or are known to be associated with phase separated membraneless organelles. Thus, our overall results present a thought-provoking hypothesis that IDRs in the R-loop interactome can provide a functional link between R-loop recognition via direct binding and downstream signaling through the assembly of LLPS-mediated membrane-less R-loop foci. The absence or dysregulation of the function of IDR-enriched R-loop interactors can potentially lead to severe genomic defects, such as the widespread R-loop-mediated DNA double strand breaks that we recently observed in Fragile X patient-derived cells.

Generalizable Framework for Algorithmic Interpretation of Thin Film Morphologies in Scanning Probe Images.

We describe an open-source and widely adaptable Python library that recognizes morphological features and domains in images collected via scanning probe microscopy. π-Conjugated polymers (CPs) are ideal for evaluating the Materials Morphology Python (m2py) library because of their wide range of morphologies and feature sizes. Using thin films of nanostructured CPs, we demonstrate the functionality of a general m2py workflow. We apply numerical methods to enhance the signals collected by the scanning probe, followed by Principal Component Analysis (PCA) to reduce the dimensionality of the data. Then, a Gaussian Mixture Model segments every pixel in the image into phases, which have similar material-property signals. Finally, the phase-labeled pixels are grouped and labeled as morphological domains using either connected components labeling or persistence watershed segmentation. These tools are adaptable to any scanning probe measurement, so the labels that m2py generates will allow researchers to individually address and analyze the identified domains in the image. This level of control, allows one to describe the morphology of the system using quantitative and statistical descriptors such as the size, distribution, and shape of the domains. Such descriptors will enable researchers to quantitatively track and compare differences within and between samples.