Engineering PapersSearch

SEARCH · Engineering Papers

Results for “Lipid Signalling”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 19 records

Mechanisms of cell signaling by nitric oxide and peroxynitrite: from mitochondria to MAP kinases

Many of the biological and pathological effects of nitric oxide (NO) are mediated through cell signaling pathways that are initiated by NO reacting with metalloproteins. More recently, it has been recognized that the reaction of NO with free radicals such as superoxide and the lipid peroxyl radical also has the potential to modulate redox signaling. Although it is clear that NO can exert both cytotoxic and cytoprotective actions, the focus of this overview are those reactions that could lead to protection of the cell against oxidative stress in the vasculature. This will include the induction of antioxidant defenses such as glutathione, activation of mitogen-activated protein kinases in response to blood flow, and modulation of mitochondrial function and its impact on apoptosis. Models are presented that show the increased synthesis of glutathione in response to shear stress and inhibition of cytochrome c release from mitochondria. It appears that in the vasculature NO-dependent signaling pathways are of three types: (i) those involving NO itself, leading to modulation of mitochondrial respiration and soluble guanylate cyclase; (ii) those that involve S-nitrosation, including inhibition of caspases; and (iii) autocrine signaling that involves the intracellular formation of peroxynitrite and the activation of the mitogen-activated protein kinases. Taken together, NO plays a major role in the modulation of redox cell signaling through a number of distinct pathways in a cellular setting.

Non-NASA Center

Up-regulation of phosphoinositide metabolism in tobacco cells constitutively expressing the human type I inositol polyphosphate 5-phosphatase

To evaluate the impact of suppressing inositol 1,4,5-trisphosphate (InsP(3)) in plants, tobacco (Nicotiana tabacum) cells were transformed with the human type I inositol polyphosphate 5-phosphatase (InsP 5-ptase), an enzyme which specifically hydrolyzes InsP(3). The transgenic cell lines showed a 12- to 25-fold increase in InsP 5-ptase activity in vitro and a 60% to 80% reduction in basal InsP(3) compared with wild-type cells. Stimulation with Mas-7, a synthetic analog of the wasp venom peptide mastoparan, resulted in an approximately 2-fold increase in InsP(3) in both wild-type and transgenic cells. However, even with stimulation, InsP(3) levels in the transgenic cells did not reach wild-type basal values, suggesting that InsP(3) signaling is compromised. Analysis of whole-cell lipids indicated that phosphatidylinositol 4,5-bisphosphate (PtdInsP(2)), the lipid precursor of InsP(3), was greatly reduced in the transgenic cells. In vitro assays of enzymes involved in PtdInsP(2) metabolism showed that the activity of the PtdInsP(2)-hydrolyzing enzyme phospholipase C was not significantly altered in the transgenic cells. In contrast, the activity of the plasma membrane PtdInsP 5 kinase was increased by approximately 3-fold in the transgenic cells. In vivo labeling studies revealed a greater incorporation of (32)P into PtdInsP(2) in the transgenic cells compared with the wild type, indicating that the rate of PtdInsP(2) synthesis was increased. These studies show that the constitutive expression of the human type I InsP 5-ptase in tobacco cells leads to an up-regulation of the phosphoinositide pathway and highlight the importance of PtdInsP(2) synthesis as a regulatory step in this system.

NASA Discipline Cell Biology

ExCALiBR: An Instrument for Uncovering the Origin of the Moon’s Organics.

Lipids are organic molecules of great astrobiological interest for future NASA missions to the Moon, Mars and Icy Worlds. Lipids are essential for life as we know it, and likely also required for putative extraterrestrial organisms. These robust organics can be synthesized abiotically and survive for long periods of geological time. Laboratory characterization techniques are well established but are laborious, operator dependent, and require large volumes of consumables, precluding in situ analysis. We are developing an autonomous, miniaturized fluidic system, integrating lab techniques for lipid analysis while minimizing rea-gent volumes and concentrating organics for analysis, thereby increasing signal-to-noise ratios by orders of magnitude. This system, the Extractor for Chemical Analysis of Lipid Biomarkers in Regolith (ExCALiBR), will enable future organic surveys by extracting and concentrating lipids from approximately 50 grams of regolith using a fluidic and microfluidic sample processor made of materials compatible with non-aqueous solvents required for extraction. ExCAL-iBR can be deployed on a lander or used on a crewed mission.

M B Wilhelm

Lipid biomarkers for bacterial ecosystems: studies of cultured organisms, hydrothermal environments and ancient sediments

This paper forms part of our long-term goal of using molecular structure and carbon isotopic signals preserved as hydrocarbons in ancient sediments to improve understanding of the early evolution of Earth's surface environment. We are particularly concerned with biomarkers which are informative about aerobiosis. Here, we combine bacterial biochemistry with the organic geochemistry of contemporary and ancient hydrothermal ecosystems to construct models for the nature, behaviour and preservation potential of primitive microbial communities. We use a combined molecular and isotopic approach to characterize lipids produced by cultured bacteria and test a variety of culture conditions which affect their biosynthesis. This information is then compared with lipid mixtures isolated from contemporary hot springs and evaluated for the kinds of chemical change that would accompany burial and incorporation into the sedimentary record. In this study we have shown that growth temperature does not appear to alter isotopic fractionation within the lipid classes produced by a methanotropic bacterium. We also found that cultured cyanobacteria biosynthesize diagnostic methylalkanes and dimethylalkanes with the latter only made when growing under low pCO2. In an examination of a microbial mat sample from Octopus Spring, Yellowstone National Park (USA), we could readily identify chemical structures with 13C contents which were diagnostic for the phototrophic organisms such as cyanobacteria and Chloroflexus. We could not, however, find molecular evidence for operation of a methane cycle in the particular mat samples we studied.

NASA Discipline Exobiology

Protobiological informatoin, bidirectional recognition and reverse translation

Emergence of protobiological information has been suggested by experiments in which heated mixtures of alpha-amino acids order themselves into a self limited array of thermal proteins. The polymers display selective catalytic, hormonal, and other activities. Interactions of varied cationic thermal proteins with polynucleotides indicate selective recognition in both directions. Reverse translation is partly a missing link in the molecular evolution flowsheet. The self ordering of amino acids serves conceptually as a deterministic evolutionary precursor of the modern coding mechanism. The possibility for the evolution of information at an early nontemplated protein stage is supported by findings of electrical signals from proteinoid microspheres prepared with no DNA/RNA in their history. The deposition of thermal copolyamino acids on lipid membranes in the Mueller-Rudin apparatus has here been found to produce electrical behavior like that evoked by bacterial EIM polypeptide. A new procedure is to make a film of membrane on the electrode; the results provide maximal repeatability. The principle of nonrandom biomacromolecular specificity identified by these studies in molecular evolution have been extrapolated to principles of evolution of advanced organisms.

Fox, S. W.

Cell signaling by reactive nitrogen and oxygen species in atherosclerosis

The production of reactive oxygen and nitrogen species has been implicated in atherosclerosis principally as means of damaging low-density lipoprotein that in turn initiates the accumulation of cholesterol in macrophages. The diversity of novel oxidative modifications to lipids and proteins recently identified in atherosclerotic lesions has revealed surprising complexity in the mechanisms of oxidative damage and their potential role in atherosclerosis. Oxidative or nitrosative stress does not completely consume intracellular antioxidants leading to cell death as previously thought. Rather, oxidative and nitrosative stress have a more subtle impact on the atherogenic process by modulating intracellular signaling pathways in vascular tissues to affect inflammatory cell adhesion, migration, proliferation, and differentiation. Furthermore, cellular responses can affect the production of nitric oxide, which in turn can strongly influence the nature of oxidative modifications occurring in atherosclerosis. The dynamic interactions between endogenous low concentrations of oxidants or reactive nitrogen species with intracellular signaling pathways may have a general role in processes affecting wound healing to apoptosis, which can provide novel insights into the pathogenesis of atherosclerosis.

Review, Tutorial

Apparatus and Methods for Manipulation and Optimization of Biological Systems

The invention provides systems and methods for manipulating biological systems, for example to elicit a more desired biological response from a biological sample, such as a tissue, organ, and/or a cell. In one aspect, the invention operates by efficiently searching through a large parametric space of stimuli and system parameters to manipulate, control, and optimize the response of biological samples sustained in the system. In one aspect, the systems and methods of the invention use at least one optimization algorithm to modify the actuator's control inputs for stimulation, responsive to the sensor's output of response signals. The invention can be used, e.g., to optimize any biological system, e.g., bioreactors for proteins, and the like, small molecules, polysaccharides, lipids, and the like. Another use of the apparatus and methods includes is for the discovery of key parameters in complex biological systems.

Ho, Chih-Ming

Biomarkers as tracers for life on early earth and Mars

Biomarkers in geological samples are products derived from biochemical (natural product) precursors by reductive and oxidative processes (e.g., cholestanes from cholesterol). Generally, lipids, pigments and biomembranes are preserved best over longer geological times and labile compounds such as amino acids, sugars, etc. are useful biomarkers for recent times. Thus, the detailed characterization of biomarker compositions permits the assessment of the major contributing species of extinct and/or extant life. In the case of the early Earth, work has progressed to elucidate molecular structure and carbon isotropic signals preserved in ancient sedimentary rocks. In addition, the combination of bacterial biochemistry with the organic geochemistry of contemporary and ancient hydrothermal ecosystems permits the modeling of the nature, behavior and preservation potential of primitive microbial communities. This approach uses combined molecular and isotopic analyses to characterize lipids produced by cultured bacteria (representative of ancient strains) and to test a variety of culture conditions which affect their biosynthesis. On considering Mars, the biomarkers from lipids and biopolymers would be expected to be preserved best if life flourished there during its early history (3.5-4 x 10(9) yr ago). Both oxidized and reduced products would be expected. This is based on the inferred occurrence of hydrothermal activity during that time with the concomitant preservation of biochemically-derived organic matter. Both known biomarkers (i.e., as elucidated for early terrestrial samples and for primitive terrestrial microbiota) and novel, potentially unknown compounds should be characterized.

Non-NASA Center

Diagnosis and Treatment of Neurological Disorders by Millimeter-Wave Stimulation

Increasingly, millimeter waves are being employed for telecomm, radar, and imaging applications. To date in the U.S, however, very few investigations on the impact of this radiation on biological systems at the cellular level have been undertaken. In the beginning, to examine the impact of millimeter waves on cellular processes, researchers discovered that cell membrane depolarization may be triggered by low levels of integrated power at these high frequencies. Such a situation could be used to advantage in the direct stimulation of neuronal cells for applications in neuroprosthetics and diagnosing or treating neurological disorders. An experimental system was set up to directly monitor cell response on exposure to continuous-wave, fixed-frequency, millimeter-wave radiation at low and modest power levels (0.1 to 100 safe exposure standards) between 50 and 100 GHz. Two immortalized cell lines derived from lung and neuronal tissue were transfected with green fluorescent protein (GFP) that locates on the inside of the cell membrane lipid bi-layer. Oxonol dye was added to the cell medium. When membrane depolarization occurs, the oxonal bound to the outer wall of the lipid bi-layer can penetrate close to the inner wall where the GFP resides. Under fluorescent excitation (488 nm), the normally green GFP (520 nm) optical signal quenches and gives rise to a red output when the oxonol comes close enough to the GFP to excite a fluorescence resonance energy transfer (FRET) with an output at 620 nm. The presence of a strong FRET signature upon exposures of 30 seconds to 2 minutes at 5-10 milliwatts per square centimeter RF power at 50 GHz, followed by a return to the normal 520-nm GFP signal after a few minutes indicating repolarization of the membrane, indicates that low levels of RF energy may be able to trigger non-destructive membrane depolarization without direct cell contact. Such a mechanism could be used to stimulate neuronal cells in the cortex without the need for invasive electrodes as millimeter waves penetrate skin and bone on the order of 15 mm in depth. Although 50 GHz could not readily penetrate from the outer skull to the center of the cortex, implants on the outer skull or even on the scalp could reach the outer layer of the cerebral cortex where substantial benefit could be realized from such non-contact type excitation.

Siegel, Peter H.

Spectral filtering for plant production

Both plants and animals have one general commonality in their perception of light. They both are sensitive primarily to the 400 to 700 nm wavelength portion of the electromagnetic spectrum. This is referred to as the visible spectrum for animals and as the photosynthetically active radiation (PAR) spectrum for plants. Within this portion of the spectrum, animals perceive colors. Relatively recently it has been learned that within this same spectral range plants also demonstrate varying responses at different wavelengths, somewhat analogous to the definition of various colors at specific wavelengths. Although invisible to the human eye, portions of the electromagnetic spectrum on either side of the visible range are relatively inactive photosynthetically but have been found to influence important biological functions. These portions include the ultraviolet (UV approximately equal to 280-400 nm) and the far-red (FR approximately equal to 700-800 nm). The basic photoreceptor of plants for photosynthesis is chlorophyll. It serves to capture radiant energy which combined with carbon dioxide and water produces oxygen and assimulated carbon, used for the synthesis of cell wall polysaccarides, proteins, membrane lipids and other cellular constituents. The energy and carbon building blocks of photosynthesis sustain growth of plants. On the other hand, however, there are other photoreceptors, or pigments, that function as signal transducers to provide information that controls many physiological and morphological responses of how a plant grows. Known photomorphogenic receptors include phytochrome (the red/far-red sensor in the narrow bands of 655-665 nm and 725-735 nm ranges, respectively) and 'cryptochrome' (the hypothetical UV-B sensor in the 280-320 nm range). Since the USDA team of W. L. Butler, S. B. Hendricks, H. A. Borthwick, H. A. Siegleman and K. Norris in Beltsville, MD detected by spectroscopy, extracted and identified phytochrome as a protein in the 1950's, many other investigators have found evidence of its control functions in plants. Considerably less, however, is known about the yet non-isolated cryptochrome. The information-transferring roles of photoreceptors in plants at specific spectral ranges quite naturally stimulated plant scientists and engineers to consider physically manipulating light to achieve desired physiological and morphological characteristics. One way to manipulate light is to filter it through materials that selectively transmit portions of the sun's spectrum in and near the PAR range.

Young, Roy E.

Endothelial Cell Morphology and Migration are Altered by Changes in Gravitational Fields

Many of the physiological changes of the cardiovascular system during space flight may originate from the dysfunction of basic biological mechanisms caused by microgravity. The weightlessness affects the system when blood and other fluids move to the upper body causing the heart to enlarge to handle the increased blood flow to the upper extremities and decrease circulating volume. Increase arterial pressure triggers baroreceptors which signal the brain to adjust heart rate. Hemodynarnic studies indicate that the microgravity-induced headward fluid redistribution results in various cardiovascular changes such as; alteration of vascular permeability resulting in lipid accumulation in the lumen of the vasculature and degeneration of the the vascular wall, capillary alteration with extensive endothelial invagination. Achieving a true microgravity environment in ground based studies for prolonged periods is virtually impossible. The application of vector-averaged gravity to mammalian cells using horizontal clinostat produces alterations of cellular behavior similar to those observed in microgravity. Similarly, the low shear, horizontally rotating bioreactor (originally designed by NASA) also duplicates several properties of microgravity. Additionally, increasing gravity, i.e., hypcrgravity is easily achieved. Hypergravity has been found to increase the proliferation of several different cell lines (e.g., chick embryo fibroblasts) while decreasing cell motility and slowing liver regeneration following partial hepatectomy. The effect of altered gravity on cells maybe similar to those of other physical forces, i.e. shear stress. Previous studies examining laminar flow and shear stress on endothelial cells found that the cells elongate, orient with the direction of flow, and reorganize their F-actin structure, with concomitant increase in cell stiffness. These studies suggest that alterations in the gravity environment will change the behavior of most cells, including vascular cells. However, few studies have been directed at assessing the effect of altered gravitational field on vascular cell fiction and metabolism, Using image analysis we examined how bovine aortic endothelial cells altered their morphological characteristics and their response to a denudation injury when cells were subjected to simulated microgravity and hypergravity.

Melhado, Caroline

Apparatus and methods for manipulation and optimization of biological systems

The invention provides systems and methods for manipulating, e.g., optimizing and controlling, biological systems, e.g., for eliciting a more desired biological response of biological sample, such as a tissue, organ, and/or a cell. In one aspect, systems and methods of the invention operate by efficiently searching through a large parametric space of stimuli and system parameters to manipulate, control, and optimize the response of biological samples sustained in the system, e.g., a bioreactor. In alternative aspects, systems include a device for sustaining cells or tissue samples, one or more actuators for stimulating the samples via biochemical, electromagnetic, thermal, mechanical, and/or optical stimulation, one or more sensors for measuring a biological response signal of the samples resulting from the stimulation of the sample. In one aspect, the systems and methods of the invention use at least one optimization algorithm to modify the actuator's control inputs for stimulation, responsive to the sensor's output of response signals. The compositions and methods of the invention can be used, e.g., to for systems optimization of any biological manufacturing or experimental system, e.g., bioreactors for proteins, e.g., therapeutic proteins, polypeptides or peptides for vaccines, and the like, small molecules (e.g., antibiotics), polysaccharides, lipids, and the like. Another use of the apparatus and methods includes combination drug therapy, e.g. optimal drug cocktail, directed cell proliferations and differentiations, e.g. in tissue engineering, e.g. neural progenitor cells differentiation, and discovery of key parameters in complex biological systems.

Ho, Chih-Ming

Subcellular targeting of nine calcium-dependent protein kinase isoforms from Arabidopsis

Calcium-dependent protein kinases (CDPKs) are specific to plants and some protists. Their activation by calcium makes them important switches for the transduction of intracellular calcium signals. Here, we identify the subcellular targeting potentials for nine CDPK isoforms from Arabidopsis, as determined by expression of green fluorescent protein (GFP) fusions in transgenic plants. Subcellular locations were determined by fluorescence microscopy in cells near the root tip. Isoforms AtCPK3-GFP and AtCPK4-GFP showed a nuclear and cytosolic distribution similar to that of free GFP. Membrane fractionation experiments confirmed that these isoforms were primarily soluble. A membrane association was observed for AtCPKs 1, 7, 8, 9, 16, 21, and 28, based on imaging and membrane fractionation experiments. This correlates with the presence of potential N-terminal acylation sites, consistent with acylation as an important factor in membrane association. All but one of the membrane-associated isoforms targeted exclusively to the plasma membrane. The exception was AtCPK1-GFP, which targeted to peroxisomes, as determined by covisualization with a peroxisome marker. Peroxisome targeting of AtCPK1-GFP was disrupted by a deletion of two potential N-terminal acylation sites. The observation of a peroxisome-located CDPK suggests a mechanism for calcium regulation of peroxisomal functions involved in oxidative stress and lipid metabolism.

Non-NASA Center

Permeation of membranes by the neutral form of amino acids and peptides: relevance to the origin of peptide translocation

The flux of amino acids and other nutrient solutes such as phosphate across lipid bilayers (liposomes) is 10(5) slower than facilitated inward transport across biological membranes. This suggest that primitive cells lacking highly evolved transport systems would have difficulty transporting sufficient nutrients for cell growth to occur. There are two possible ways by which early life may have overcome this difficulty: (1) The membranes of the earliest cellular life-forms may have been intrinsically more permeable to solutes; or (2) some transport mechanism may have been available to facilitate transbilayer movement of solutes essential for cell survival and growth prior to the evolution of membrane transport proteins. Translocation of neutral species represents one such mechanism. The neutral forms of amino acids modified by methylation (creating protonated weak bases) permeate membranes up to 10(10) times faster than charged forms. This increased permeability when coupled to a transmembrane pH gradient can result in significantly increased rates of net unidirectional transport. Such pH gradients can be generated in vesicles used to model protocells that preceded and were presumably ancestral to early forms of life. This transport mechanism may still play a role in some protein translocation processes (e.g. for certain signal sequences, toxins and thylakoid proteins) in vivo.

NASA Discipline Exobiology

Transcriptomics-based Machine Learning Analysis Predicts Space-Exposed Murine Livers

Limited sample sizes, high data dimensionality, and sensitivity to technical and biological variability of next generation sequencing (NGS), has typically limited machine learning (ML) in space studies and further study of radiation effects. However, pooling smaller studies while addressing intra- and inter-study variabilities allows for ML predictive modeling. Here, integration methods were applied to whole transcriptome shotgun sequencing (RNAseq) data from 6 mouse liver GeneLab datasets (GLDS) with a total of 113 spaceflight and ground-control samples to determine top features relevant to spaceflight including the effect of radiation exposure. Data was normalized within each study, then merged and scaled across all datasets. Data dimensionality was reduced using a minimum redundancy maximum relevance (MRMR) methodology. The top MRMR features were used to predict spaceflight vs. ground-control samples using a Random Forest (RF) classifier with 5-fold cross validation (CV). The ML-based gene sets were further compared against differential gene expression results from individual GLDS. CV training using the top 100 MRMR genes show averages of 86% accuracy and 0.95 AUC value on the validation set over 5 folds (Figure 1A). Baseline set analysis on differentially expressed genes (DEGs) identified using padj ≤ 0.05 show 811 or 68 DEGs overlapping between at least 2 or 3 studies, respectively (Figure 1B). Over-representation analysis showed overlapping biological processes related to fatty acid and lipid metabolism. Set analysis between the MRMR features and the DEGs showed 60 or 8 genes overlapping with at least 1 or 2 studies, respectively. MRMR feature selection and ensemble ML methods (e.g. RF) improve performance relative to a Naïve Bayes classifier when NGS data sets are analyzed. A challenge of applying ML methods across heterogeneous NGS data is accounting for signal:noise ratio. Here, signal validation across studies was shown by intersecting sets between top MRMR genes and DEGs from RNASeq analysis. Non-intersecting sets introduce opportunity to explore spaceflight relevant genes and implementing ML methods across existing NGS datasets may overcome sample size limitations. ML coupled with existing analytical methods enhances understanding of disease by revealing common underlying pathways across datasets.

Machine Learning

Transcriptomics-based Machine Learning Analysis Predicts Space-Exposed Murine Livers

Limited sample sizes, high data dimensionality, and sensitivity to technical and biological variability of next generation sequencing (NGS), typically limits machine learning (ML) approaches in spaceflight studies that include radiation effects. However, pooling smaller studies while addressing intra- and inter-study variabilities allows for ML predictive modeling. Here, integration methods were applied to whole transcriptome shotgun sequencing (RNA-seq) data from six mouse liver GeneLab datasets (GLDS) (n ranging from 6 to 39 samples) from with a total of 81 spaceflight and ground-control samples to determine top features (i.e. genes) relevant to spaceflight including the effect of radiation exposure. RNASeq counts were normalized for each study, then merged and scaled across all datasets. Data dimensionality was reduced using a minimum redundancy maximum relevance (MRMR) methodology. Redundancy and relevance were computed using the Pearson correlation and F-statistic, respectively. The top 100 MRMR features were used to predict spaceflight vs. ground-control samples using Random Forest (RF), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA) classifiers with 5-fold cross validation (CV). Principal component analysis (PCA) on the complete feature set versus the MRMR features shows separation between spaceflight samples and ground controls (Figure 1A). The ML-based gene sets were compared against differential gene expression results obtained with DESeq2 from individual GLDS. Using all features or randomly sampled subsets at matching set sizes with MRMR, a maximum classifier accuracy of 69% was shown on the test set over 5 folds. For all classifiers, CV training using at least the top 30 MRMR genes show minimum 89% accuracy and 0.95 AUC value on the test set over 5 folds (Figure 1B). Baseline set analysis on differentially expressed genes (DEGs) identified using padj ≤ 0.05 show 295 DEGs that overlap at least two studies and 13 DEGs that overlap three studies (Figure 1C). Set analysis between the top 100 MRMR features and the DEGs showed 47 genes that overlap at least one study and 24 genes that overlap two studies. Over-representation analysis showed overlapping biological processes related to fatty acid and lipid metabolism which may indicate these processes in the response to spaceflight stressors. MRMR feature selection for the selected ML methods improve performance relative to a classifier built on all features or randomly sampled subsets. Permutation feature importance within the decorrelated MRMR features showed concordance in feature ranking between ML methods. A challenge of applying ML methods across heterogeneous NGS data is accounting for signal:noise. Here, signal validation across studies was shown by intersecting sets between top MRMR genes and DEGs from DESeq2 analysis. Non-intersecting sets introduce opportunity to explore genes relevant to differentiating space flight exposed groups and implementing ML methods across existing NGS datasets may overcome sample size limitations.

Machine Learning

Transcriptomics-based Machine Learning (ML) Analysis Predicts Space-Exposed Murine Livers

Limited sample sizes, high data dimensionality, and sensitivity to technical and biological variability of next generation sequencing (NGS), typically limits machine learning (ML) approaches in spaceflight studies that include radiation effects. However, pooling smaller studies while addressing intra- and inter-study variabilities allows for ML predictive modeling. Here, integration methods were applied to whole transcriptome shotgun sequencing (RNA-seq) data from six mouse liver GeneLab datasets (GLDS) (n ranging from 6 to 39 samples) from with a total of 81 spaceflight and ground-control samples to determine top features (i.e. genes) relevant to spaceflight including the effect of radiation exposure. RNASeq counts were normalized for each study, then merged and scaled across all datasets. Data dimensionality was reduced using a minimum redundancy maximum relevance (MRMR) methodology. Redundancy and relevance were computed using the Pearson correlation and F-statistic, respectively. The top 100 MRMR features were used to predict spaceflight vs. ground-control samples using Random Forest (RF), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA) classifiers with 5-fold cross validation (CV). Principal component analysis (PCA) on the complete feature set versus the MRMR features shows separation between spaceflight samples and ground controls (Figure 1A). The ML-based gene sets were compared against differential gene expression results obtained with DESeq2 from individual GLDS. Using all features or randomly sampled subsets at matching set sizes with MRMR, a maximum classifier accuracy of 69% on the test set over 5 folds. For all classifiers, CV training using at least the top 30 MRMR genes show minimum 89% accuracy and 0.95 AUC value on the test set over 5 folds (Figure 1B). Baseline set analysis on differentially expressed genes (DEGs) identified using padj ≤ 0.05 show 295 DEGs that overlap at least two studies and 13 DEGs that overlap three studies (Figure 1C). Set analysis between the top 100 MRMR features and the DEGs showed 47 genes that overlap at least one study and 24 genes that overlap two studies. Over-representation analysis showed overlapping biological processes related to fatty acid and lipid metabolism which may indicate these processes in the response to spaceflight stressors. MRMR feature selection for the selected ML methods improve performance relative to a classifier built on all features or randomly sampled subsets. Permutation feature importance within the decorrelated MRMR features showed concordance in feature ranking between ML methods. A challenge of applying ML methods across heterogeneous NGS data is accounting for signal:noise. Here, signal validation across studies was shown by intersecting sets between top MRMR genes and DEGs from DESeq2 analysis. Non-intersecting sets introduce opportunity to explore genes relevant to differentiating space flight exposed groups and implementing ML methods across existing NGS datasets may overcome sample size limitations.

Machine Learning

Transcriptomics-based Machine Learning Analysis Predicts Space-Exposed Murine Livers

Limited sample sizes, high data dimensionality, and sensitivity to technical and biological variability of next generation sequencing (NGS), typically limits machine learning (ML) approaches in spaceflight studies that include radiation effects. However, pooling smaller studies while addressing intra- and inter-study variabilities allows for ML predictive modeling. Here, integration methods were applied to whole transcriptome shotgun sequencing (RNA-seq) data from six mouse liver GeneLab datasets (GLDS) (n ranging from 6 to 39 samples) from with a total of 81 spaceflight and ground-control samples to determine top features (i.e. genes) relevant to spaceflight including the effect of radiation exposure. RNASeq counts were normalized for each study, then merged and scaled across all datasets. Data dimensionality was reduced using a minimum redundancy maximum relevance (MRMR) methodology. Redundancy and relevance were computed using the Pearson correlation and F-statistic, respectively. The top 100 MRMR features were used to predict spaceflight vs. ground-control samples using Random Forest (RF), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA) classifiers with 5-fold cross validation (CV). Principal component analysis (PCA) on the complete feature set versus the MRMR features shows separation between spaceflight samples and ground controls (Figure 1A). The ML-based gene sets were compared against differential gene expression results obtained with DESeq2 from individual GLDS. Using all features or randomly sampled subsets at matching set sizes with MRMR, a maximum classifier accuracy of 69% was shown on the test set over 5 folds. For all classifiers, CV training using at least the top 30 MRMR genes show minimum 89% accuracy and 0.95 AUC value on the test set over 5 folds (Figure 1B). Baseline set analysis on differentially expressed genes (DEGs) identified using padj ≤ 0.05 show 295 DEGs that overlap at least two studies and 13 DEGs that overlap three studies (Figure 1C). Set analysis between the top 100 MRMR features and the DEGs showed 47 genes that overlap at least one study and 24 genes that overlap two studies. Over-representation analysis showed overlapping biological processes related to fatty acid and lipid metabolism which may indicate these processes in the response to spaceflight stressors. MRMR feature selection for the selected ML methods improve performance relative to a classifier built on all features or randomly sampled subsets. Permutation feature importance within the decorrelated MRMR features showed concordance in feature ranking between ML methods. A challenge of applying ML methods across heterogeneous NGS data is accounting for signal:noise. Here, signal validation across studies was shown by intersecting sets between top MRMR genes and DEGs from DESeq2 analysis. Non-intersecting sets introduce opportunity to explore genes relevant to differentiating space flight exposed groups and implementing ML methods across existing NGS datasets may overcome sample size limitations.

Machine Learning