Precision medical image hash retrieval by interpretability and feature fusion.

BACKGROUND AND OBJECTIVE: To address the problem of low accuracy of medical image retrieval due to high inter-class similarity and easy omission of lesions, a precision medical image hash retrieval method combining interpretability and feature fusion is proposed, taking chest X-ray images as an example. METHODS: Firstly, the DenseNet-121 network is pre-trained on a large dataset of medical images without manual annotation using the comparison to learn (C2L) method to obtain a backbone network model containing more medical representations with training weights. Then, a global network is constructed by using global image learning to acquire an interpretable saliency map as attention mechanisms, which can generate a mask crop to get a local discriminant region. Thirdly, the local discriminant regions are used as local network inputs to obtain local features, and the global features are used with the local features by dimension in the pooling layer. Finally, a hash layer is added between the fully connected layer and the classification layer of the backbone network, defining classification loss, quantization loss and bit-balanced loss functions to generate high-quality hash codes. The final retrieval result is output by calculating the similarity metric of the hash codes. RESULTS: Experiments on the Chest X-ray8 dataset demonstrate that our proposed interpretable saliency map can effectively locate focal regions, the fusion of features can avoid information omission, and the combination of three loss functions can generate more accurate hash codes. Compared with the current advanced medical image retrieval methods, this method can effectively improve the accuracy of medical image retrieval. CONCLUSIONS: The proposed hash retrieval approach combining interpretability and feature fusion can effectively improve the accuracy of medical image retrieval which can be potentially applied in computer-aided-diagnosis systems.

The effect of LRRK2 loss-of-function variants in humans.

Human genetic variants predicted to cause loss-of-function of protein-coding genes (pLoF variants) provide natural in vivo models of human gene inactivation and can be valuable indicators of gene function and the potential toxicity of therapeutic inhibitors targeting these genes1,2. Gain-of-kinase-function variants in LRRK2 are known to significantly increase the risk of Parkinson's disease3,4, suggesting that inhibition of LRRK2 kinase activity is a promising therapeutic strategy. While preclinical studies in model organisms have raised some on-target toxicity concerns5-8, the biological consequences of LRRK2 inhibition have not been well characterized in humans. Here, we systematically analyze pLoF variants in LRRK2 observed across 141,456 individuals sequenced in the Genome Aggregation Database (gnomAD)9, 49,960 exome-sequenced individuals from the UK Biobank and over 4 million participants in the 23andMe genotyped dataset. After stringent variant curation, we identify 1,455 individuals with high-confidence pLoF variants in LRRK2. Experimental validation of three variants, combined with previous work10, confirmed reduced protein levels in 82.5% of our cohort. We show that heterozygous pLoF variants in LRRK2 reduce LRRK2 protein levels but that these are not strongly associated with any specific phenotype or disease state. Our results demonstrate the value of large-scale genomic databases and phenotyping of human loss-of-function carriers for target validation in drug discovery.

Optical coherence tomography modeling incorporating scattering, absorption, and multiple reflections.

A direct scattering optical coherence tomography forward model was developed to simulate A-scans for both idealized and real light sources on an arbitrary given sample structure. Previous models neglected absorption, scattering, and multiple reflections at interfacial layers, and so two extended models were developed to investigate the impact of these processes. The first model uses the Beer-Lambert law to incorporate both absorption and scattering optical processes, and the second model uses a recursive form to model multiple reflections. These models were tested on a structure representative of a multilayered skin sample. The results show that the absorption and scattering processes have significant impact on the height of the peaks in the simulated A-scans. Conversely, the incorporation of multiple reflections has very little impact on the height of these peaks. Neither of the above processes has any impact on the locations of the A-scan peaks, which are associated with the sample interfaces between layers.

Biochemical Evidence for a Nuclear Modifier Allele (A10S) in TRMU (Methylaminomethyl-2-thiouridylate-methyltransferase) Related to Mitochondrial tRNA Modification in the Phenotypic Manifestation of Deafness-associated 12S rRNA Mutation.

Nuclear modifier gene(s) was proposed to modulate the phenotypic expression of mitochondrial DNA mutation(s). Our previous investigations revealed that a nuclear modifier allele (A10S) in TRMU (methylaminomethyl-2-thiouridylate-methyltransferase) related to tRNA modification interacts with 12S rRNA 1555A→G mutation to cause deafness. The A10S mutation resided at a highly conserved residue of the N-terminal sequence. It was hypothesized that the A10S mutation altered the structure and function of TRMU, thereby causing mitochondrial dysfunction. Using molecular dynamics simulations, we showed that the A10S mutation introduced the Ser10 dynamic electrostatic interaction with the Lys106 residue of helix 4 within the catalytic domain of TRMU. The Western blotting analysis displayed the reduced levels of TRMU in mutant cells carrying the A10S mutation. The thermal shift assay revealed the Tm value of mutant TRMU protein, lower than that of the wild-type counterpart. The A10S mutation caused marked decreases in 2-thiouridine modification of U34 of tRNALys, tRNAGlu and tRNAGln However, the A10S mutation mildly increased the aminoacylated efficiency of tRNAs. The altered 2-thiouridine modification worsened the impairment of mitochondrial translation associated with the m.1555A→G mutation. The defective translation resulted in the reduced activities of mitochondrial respiration chains. The respiratory deficiency caused the reduction of mitochondrial ATP production and elevated the production of reactive oxidative species. As a result, mutated TRMU worsened mitochondrial dysfunctions associated with m.1555A→G mutation, exceeding the threshold for expressing a deafness phenotype. Our findings provided new insights into the pathophysiology of maternally inherited deafness that was manifested by interaction between mtDNA mutation and nuclear modifier gene.

The functional significance of common polymorphisms in zinc finger transcription factors.

Variants that alter the DNA-binding specificity of transcription factors could affect the specificity for and expression of potentially many target genes, as has been observed in several tumor-derived mutations. Here we examined if such trans expression quantitative trait loci (trans-eQTLs) could similarly result from common genetic variants. We chose to focus on the Cys2-His2 class of zinc finger transcription factors because they are the most abundant superfamily of transcription factors in human and have well-characterized DNA binding interactions. We identified 430 SNPs that cause missense substitutions in the DNA-contacting residues. Fewer common missense SNPs were found at DNA-contacting residues compared with non-DNA-contacting residues (P = 0.00006), consistent with possible functional selection against SNPs at DNA-contacting positions. Functional predictions based on zinc finger transcription factor (ZNF) DNA binding preferences also suggested that many common substitutions could potentially alter binding specificity. However, Hardy-Weinberg Equilibrium analysis and examination of seven orthologs within the primate lineage failed to find evidence of trans-eQTLs associated with the DNA-contacting positions or evidence of a different selection pressure on a contemporary and evolutionary timescales. The overall conclusion was that common SNPs that alter the DNA-contacting residues of these factors are unlikely to produce strong trans-eQTLs, consistent with the observations by others that trans-eQTLs in humans tend to be few and weak. Some rare SNPs might alter specificity and remained rare due to purifying selection. The study also underscores the need for large-scale eQTLs mapping efforts that might provide experimental evidence for SNPs that alter the choice of transcription factor binding sites.

A genome-wide siRNA screen reveals diverse cellular processes and pathways that mediate genome stability.

Signaling pathways that respond to DNA damage are essential for the maintenance of genome stability and are linked to many diseases, including cancer. Here, a genome-wide siRNA screen was employed to identify additional genes involved in genome stabilization by monitoring phosphorylation of the histone variant H2AX, an early mark of DNA damage. We identified hundreds of genes whose downregulation led to elevated levels of H2AX phosphorylation (gammaH2AX) and revealed links to cellular complexes and to genes with unclassified functions. We demonstrate a widespread role for mRNA-processing factors in preventing DNA damage, which in some cases is caused by aberrant RNA-DNA structures. Furthermore, we connect increased gammaH2AX levels to the neurological disorder Charcot-Marie-Tooth (CMT) syndrome, and we find a role for several CMT proteins in the DNA-damage response. These data indicate that preservation of genome stability is mediated by a larger network of biological processes than previously appreciated.