On the interpretability of machine learning-based model for predicting hypertension

BACKGROUND

Although complex machine learning models are commonly outperforming the traditional simple interpretable models, clinicians find it hard to understand and trust these complex models due to the lack of intuition and explanation of their predictions. The aim of this study to demonstrate the utility of various model-agnostic explanation techniques of machine learning models with a case study for analyzing the outcomes of the machine learning random forest model for predicting the individuals at risk of developing hypertension based on cardiorespiratory fitness data.

METHODS

The dataset used in this study contains information of 23,095 patients who underwent clinician-referred exercise treadmill stress testing at Henry Ford Health Systems between 1991 and 2009 and had a complete 10-year follow-up. Five global interpretability techniques (Feature Importance, Partial Dependence Plot, Individual Conditional Expectation, Feature Interaction, Global Surrogate Models) and two local interpretability techniques (Local Surrogate Models, Shapley Value) have been applied to present the role of the interpretability techniques on assisting the clinical staff to get better understanding and more trust of the outcomes of the machine learning-based predictions.

RESULTS

Several experiments have been conducted and reported. The results show that different interpretability techniques can shed light on different insights on the model behavior where global interpretations can enable clinicians to understand the entire conditional distribution modeled by the trained response function. In contrast, local interpretations promote the understanding of small parts of the conditional distribution for specific instances.

CONCLUSIONS

Various interpretability techniques can vary in their explanations for the behavior of the machine learning model. The global interpretability techniques have the advantage that it can generalize over the entire population while local interpretability techniques focus on giving explanations at the level of instances. Both methods can be equally valid depending on the application need. Both methods are effective methods for assisting clinicians on the medical decision process, however, the clinicians will always remain to hold the final say on accepting or rejecting the outcome of the machine learning models and their explanations based on their domain expertise.

New lupeol esters as active substances in the treatment of skin damage

The purpose of the research was to obtain new derivatives of natural triterpene lupeol and to evaluate their potential as active substances in the treatment of skin damage. Four new lupeol esters (propionate, succinate, isonicotinate and acetylsalicylate) and lupeol acetate were obtained using an eco-friendly synthesis method. In the esterification process, the commonly used hazardous reagents in this type of synthesis were replaced by safe ones. This unconventional, eco-friendly, method is particularly important because the compounds obtained are potentially active substances in skin care formulations. Even trace amounts of hazardous reagents can have a toxic effect on damaged or irritated tissues. The molecular structure of the esters were confirmed by ¹H NMR, ¹³C NMR and IR spectroscopy methods. Their crystal structures were determined using XRD method. To complete the analysis of their characteristics, physicochemical properties (melting point, lipophilicity, water solubility) and biological activity of the lupeol derivatives were studied. Results of an irritant potential test, carried out on Reconstructed Human Epidermis (RHE), confirmed that the synthesized lupeol derivatives are not cytotoxic and they stimulate a process of human cell proliferation. The safety of use for tested compounds was determined in a cell viability test (cytotoxicity detection kit based on the measurement of lactate dehydrogenase activity) for keratinocytes and fibroblasts. The results obtained showed that the modification of lupeol structure improve its bioavailability and activity. All of the esters penetrate the stratum corneum and the upper layers of the dermis better than the maternal lupeol. Lupeol isonicotinate, acetate and propionate were the most effective compounds in a stimulation of the human skin cell proliferation process. This combination resulted in an increase in the concentration of cells of more than 30% in comparison to control samples. The results indicate that the chemical modification of lupeol allows to obtain promising active substances for treatment of skin damage, including thermal, chemical and radiation burns.

Choice of assembly software has a critical impact on virome characterisation

BACKGROUND

The viral component of microbial communities plays a vital role in driving bacterial diversity, facilitating nutrient turnover and shaping community composition. Despite their importance, the vast majority of viral sequences are poorly annotated and share little or no homology to reference databases. As a result, investigation of the viral metagenome (virome) relies heavily on de novo assembly of short sequencing reads to recover compositional and functional information. Metagenomic assembly is particularly challenging for virome data, often resulting in fragmented assemblies and poor recovery of viral community members. Despite the essential role of assembly in virome analysis and difficulties posed by these data, current assembly comparisons have been limited to subsections of virome studies or bacterial datasets.

DESIGN

This study presents the most comprehensive virome assembly comparison to date, featuring 16 metagenomic assembly approaches which have featured in human virome studies. Assemblers were assessed using four independent virome datasets, namely, simulated reads, two mock communities, viromes spiked with a known phage and human gut viromes.

RESULTS

Assembly performance varied significantly across all test datasets, with SPAdes (meta) performing consistently well. Performance of MIRA and VICUNA varied, highlighting the importance of using a range of datasets when comparing assembly programs. It was also found that while some assemblers addressed the challenges of virome data better than others, all assemblers had limitations. Low read coverage and genomic repeats resulted in assemblies with poor genome recovery, high degrees of fragmentation and low-accuracy contigs across all assemblers. These limitations must be considered when setting thresholds for downstream analysis and when drawing conclusions from virome data.

Study of fast and slow consecutive processes by heterogeneous isotope exchange using P-32 radiotracer

The sorption of phosphate on soils is studied by radioisotopic tracer method. Two consecutive processes with rather different rates were differentiated: namely the heterogeneous isotope exchange between the phosphate in the soil solution and the weakly sorbed phosphate (fast reaction), and the transformation of weakly sorbed phosphate to tightly sorbed phosphate (slow reaction). In this paper, it is shown how the rate constants of these two processes can be determined by a radiotracer with a relatively short half-life.

Evolution of mitochondrial TAT translocases illustrates the loss of bacterial protein transport machines in mitochondria

BACKGROUND

Bacteria and mitochondria contain translocases that function to transport proteins across or insert proteins into their inner and outer membranes. Extant mitochondria retain some bacterial-derived translocases but have lost others. While BamA and YidC were integrated into general mitochondrial protein transport pathways (as Sam50 and Oxa1), the inner membrane TAT translocase, which uniquely transports folded proteins across the membrane, was retained sporadically across the eukaryote tree.

RESULTS

We have identified mitochondrial TAT machinery in diverse eukaryotic lineages and define three different types of eukaryote-encoded TatABC-derived machineries (TatAC, TatBC and TatC-only). Here, we investigate TatAC and TatC-only machineries, which have not been studied previously. We show that mitochondria-encoded TatAC of the jakobid Andalucia godoyi represent the minimal functional pathway capable of substituting for the Escherichia coli TatABC complex and can transport at least one substrate. However, selected TatC-only machineries, from multiple eukaryotic lineages, were not capable of supporting the translocation of this substrate across the bacterial membrane. Despite the multiple losses of the TatC gene from the mitochondrial genome, the gene was never transferred to the cell nucleus. Although the major constraint preventing nuclear transfer of mitochondrial TatC is likely its high hydrophobicity, we show that in chloroplasts, such transfer of TatC was made possible due to modifications of the first transmembrane domain.

CONCLUSIONS

At its origin, mitochondria inherited three inner membrane translocases Sec, TAT and Oxa1 (YidC) from its bacterial ancestor. Our work shows for the first time that mitochondrial TAT has likely retained its unique function of transporting folded proteins at least in those few eukaryotes with TatA and TatC subunits encoded in the mitochondrial genome. However, mitochondria, in contrast to chloroplasts, abandoned the machinery multiple times in evolution. The overall lower hydrophobicity of the Oxa1 protein was likely the main reason why this translocase was nearly universally retained in mitochondrial biogenesis pathways.

Unexpected consequences of bombing. Community level response of epiphytic diatoms to environmental stress in a saline bomb crater pond area

The spatial response of epiphytic diatom communities to environmental stress was studied in a moderately saline wetland area located in the plain of Danube-Tisza Interfluve, Hungary. The area is characterised by World War II bomb crater ponds and can be regarded as an excellent ecological model system where the dispersion of species is slightly limited by distance. To study the effect of environmental variables on the communities, canonical correspondence analysis was applied. Salinity, pH, total suspended solids, total phosphorous and depth proved to be significant environmental drivers in this analysis. The ecological status of the ponds was assessed with Ziemann’s halobity index, as the trophity-depending metric cannot be applied to these habitats (due to the naturally high phosphorus content). Ponds in “good” ecological status significantly differed from those appertaining to water quality category of “not-good” ecological status considering characteristic of natural astatic soda pans (e.g. salinity, pH, ammonium, total phosphorous concentration, nitrogen:phosphorous ratio and turbidity). The differences between epiphytic diatom communities inhabiting the ponds were detected using non-parametric multidimensional scaling. The samples formed three groups according to the types of ponds (“transparent”, “transitional” and “turbid”) based on the width of the macrophyte belt around them. Indicator species related to the ecological status of the ponds and diatom communities contributing to the separation of groups of ponds were identified. One of the indicator species differed from species already described. Light and scanning electron microscopy features and phylogenetic analyses based on three genes (18S and 28S rRNA genes, rbcL) proved that it was a new species of Nitzschia genus, closely related to Nitzschia frustulum and Nitzschia inconspicua. Therefore, description of a new species, Nitzschia reskoi Ács, Duleba, C.E.Wetzel & Ector is proposed. We concluded that the increasing abundance of Nitzschia reskoi was a signal of the degradation of the intermittent saline wetlands.

Stimulating the nucleus accumbens in obesity: A positron emission tomography study after deep brain stimulation in a rodent model

PURPOSE

The nucleus accumbens (NAcc) has been suggested as a possible target for deep brain stimulation (DBS) in the treatment of obesity. Our hypothesis was that NAcc-DBS would modulate brain regions related to reward and food intake regulation, consequently reducing the food intake and, finally, the weight gain. Therefore, we examined changes in brain glucose metabolism, weight gain and food intake after NAcc-DBS in a rat model of obesity.

PROCEDURES

Electrodes were bilaterally implanted in 2 groups of obese Zucker rats targeting the NAcc. One group received stimulation one hour daily during 15 days, while the other remained as control. Weight and daily consumption of food and water were everyday registered the days of stimulation, and twice per week during the following month. Positron emission tomography (PET) studies with 2-deoxy-2-[18F]fluoro-D-glucose (FDG) were performed 1 day after the end of DBS. PET data was assessed by statistical parametric mapping (SPM12) software and region of interest (ROI) analyses.

RESULTS

NAcc-DBS lead to increased metabolism in the cingulate-retrosplenial-parietal association cortices, and decreased metabolism in the NAcc, thalamic and pretectal nuclei. Furthermore, ROIs analyses confirmed these results by showing a significant striatal and thalamic hypometabolism, and a cortical hypermetabolic region. However, NAcc-DBS did not induce a decrease in either weight gain or food intake.

CONCLUSIONS

NAcc-DBS led to changes in the metabolism of regions associated with cognitive and reward systems, whose impairment has been described in obesity.

Reconstruction of a genome-scale metabolic model for Actinobacillus succinogenes 130Z

BACKGROUND

Actinobacillus succinogenes is a promising bacterial catalyst for the bioproduction of succinic acid from low-cost raw materials. In this work, a genome-scale metabolic model was reconstructed and used to assess the metabolic capabilities of this microorganism under producing conditions.

RESULTS

The model, iBP722, was reconstructed based on the functional reannotation of the complete genome sequence of A. succinogenes 130Z and manual inspection of metabolic pathways, covering 1072 enzymatic reactions associated with 722 metabolic genes that involve 713 metabolites. The highly curated model was effective in capturing the growth of A. succinogenes on various carbon sources, as well as the SA production under various growth conditions with fair agreement between experimental and predicted data. Calculated flux distributions under different conditions show that a number of metabolic pathways are affected by the activity of some metabolic enzymes at key nodes in metabolism, including the transport mechanism of carbon sources and the ability to fix carbon dioxide.

CONCLUSIONS

The established genome-scale metabolic model can be used for model-driven strain design and medium alteration to improve succinic acid yields.

Identification of novel substrates of Shigella T3SA through analysis of its virulence plasmid-encoded secretome

Many human Gram-negative bacterial pathogens express a Type Three Secretion Apparatus (T3SA), including among the most notorious Shigella spp., Salmonella enterica, Yersinia enterocolitica and enteropathogenic Escherichia coli (EPEC). These bacteria express on their surface multiple copies of the T3SA that mediate the delivery into host cells of specific protein substrates critical to pathogenesis. Shigella spp. are Gram-negative bacterial pathogens responsible for human bacillary dysentery. The effector function of several Shigella T3SA substrates has largely been studied but their potential cellular targets are far from having been comprehensively delineated. In addition, it is likely that some T3SA substrates have escaped scrutiny as yet. Indeed, sequencing of the virulence plasmid of Shigella flexneri has revealed numerous open reading frames with unknown functions that could encode additional T3SA substrates. Taking advantage of label-free mass spectrometry detection of proteins secreted by a constitutively secreting strain of S. flexneri, we identified five novel substrates of the T3SA. We further confirmed their secretion through the T3SA and translocation into host cells using β-lactamase assays. The coding sequences of two of these novel T3SA substrates (Orf13 and Orf131a) have a guanine-cytosine content comparable to those of T3SA components and effectors. The three other T3SA substrates identified (Orf48, Orf86 and Orf176) have significant homology with antitoxin moieties of type II Toxin-Antitoxin systems usually implicated in the maintenance of low copy plasmids. While Orf13 and Orf131a might constitute new virulence effectors contributing to S. flexneri pathogenicity, potential roles for the translocation into host cells of antitoxins or antitoxin-like proteins during Shigella infection are discussed.