Towards Generalizability and Robustness in Biological Object Detection in Electron Microscopy Images.

Machine learning approaches have the potential for meaningful impact in the biomedical field. However, there are often challenges unique to biomedical data that prohibits the adoption of these innovations. For example, limited data, data volatility, and data shifts all compromise model robustness and generalizability. Without proper tuning and data management, deploying machine learning models in the presence of unaccounted for corruptions leads to reduced or misleading performance. This study explores techniques to enhance model generalizability through iterative adjustments. Specifically, we investigate a detection tasks using electron microscopy images and compare models trained with different normalization and augmentation techniques. We found that models trained with Group Normalization or texture data augmentation outperform other normalization techniques and classical data augmentation, enabling them to learn more generalized features. These improvements persist even when models are trained and tested on disjoint datasets acquired through diverse data acquisition protocols. Results hold true for transformerand convolution-based detection architectures. The experiments show an impressive 29% boost in average precision, indicating significant enhancements in the model's generalizibality. This underscores the models' capacity to effectively adapt to diverse datasets and demonstrates their increased resilience in real-world applications.

Insights into Lou Gehrig's disease from the structure and instability of the A4V mutant of human Cu,Zn superoxide dismutase.

Mutations in human superoxide dismutase (HSOD) have been linked to the familial form of amyotrophic lateral sclerosis (FALS). Amyotrophic lateral sclerosis (ALS or Lou Gehrig's disease) is one of the most common neurodegenerative disorders in humans. In ALS patients, selective killing of motor neurons leads to progressive paralysis and death within one to five years of onset. The most frequent FALS mutation in HSOD, Ala4-->Val, is associated with the most rapid disease progression. Here we identify and characterize key differences in the stability between the A4V mutant protein and its thermostable parent (HSOD-AS), in which free cysteine residues were mutated to eliminate interferences from cysteine oxidation. Denaturation studies reveal that A4V unfolds at a guanidine-HCl concentration 1M lower than HSOD-AS, revealing that A4V is significantly less stable than HSOD-AS. Determination and analysis of the crystallographic structures of A4V and HSOD-AS reveal structural features likely responsible for the loss of architectural stability of A4V observed in the denaturation experiments. The combined structural and biophysical results presented here argue that architectural destabilization of the HSOD protein may underlie the toxic function of the many HSOD FALS mutations.

ALS mutants of human superoxide dismutase form fibrous aggregates via framework destabilization.

Many point mutations in human Cu,Zn superoxide dismutase (SOD) cause familial amyotrophic lateral sclerosis (FALS), a fatal neurodegenerative disorder in heterozygotes. Here we show that these mutations cluster in protein regions influencing architectural integrity. Furthermore, crystal structures of SOD wild-type and FALS mutant H43R proteins uncover resulting local framework defects. Characterizations of beta-barrel (H43R) and dimer interface (A4V) FALS mutants reveal reduced stability and drastically increased aggregation propensity. Moreover, electron and atomic force microscopy indicate that these defects promote the formation of filamentous aggregates. The filaments resemble those seen in neurons of FALS patients and bind both Congo red and thioflavin T, suggesting the presence of amyloid-like, stacked beta-sheet interactions. These results support free-cysteine-independent aggregation of FALS mutant SOD as an integral part of FALS pathology. They furthermore provide a molecular basis for the single FALS disease phenotype resulting from mutations of diverse side-chains throughout the protein: many FALS mutations reduce structural integrity, lowering the energy barrier for fibrous aggregation.