Search NASA⌕ Search

SEARCH · Search NASA

Results for “Discriminant Analysis”

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 145 records · Page 8

Chemical studies of H chondrites. 4: New data and comparison of Antarctic suites

We report data for the trace elements Au, Co, Sb, Ga, Rb, Ag, Se, Cs, Te, Zn, Cd, Bi, Ti, and In (ordered by putative volatility during nebular condensation and accretion) determined by neutron activation analysis in 13 H5 chondrites from Victoria Land and 20 H4-6 chondrites from Queen Maud Land, Antarctica. These and earlier results provide Antarctic sample suites of 34 chondrites from Victoria Land and 25 from Queen Maud Land. Treatment of data for the most volatile 10 elements (Rb to In) in these studies by multivariate statistical techniques more robust, as well as more conservative, than conventional linear discriminant analysis and logistic regression demonstrates that compositions differ at marginally significant levels. This difference cannot be explained by trivial (terrestrial) causes and becomes more significant, despite the smaller size of the database, when comparisons are limited to data from a single analyst and when all upper limits are eliminated from consideration. The Victoria Land and Queen Maud Land suites have different mean terrestrial ages (approximately 300 kyr and approximately 100 kyr, respectively) and age distributions, suggesting that a time-dependent variation of chondritic sources with different thermal histories is responsible. As a result, these two Antarctic suites are, on average, chemically distinguishable from each other. Since H chondrites serve as a paradigm for other meteorite classes, these results indicate that the near-Earth populations of planetary materials varied with time on the 10(exp 5)-year timescale.

Wolf, Stephen F.↗

Hyperspectral Imaging and SPA–LDA Quantitative Analysis for Detection of Colon Cancer Tissue

Hyperspectral imaging (HSI) has been demonstrated to provide a rapid, precise, and noninvasive method for cancer detection. However, because HSI contains many data, quantitative analysis is often necessary to distill information useful for distinguishing cancerous from normal tissue. To demonstrate that HSI with our proposed algorithm can make this distinction, we built a Vis–NIR HSI setup and made many spectral images of colon tissues, and then used a successive projection algorithm (SPA) to analyze the hyperspectral image data of the tissues. This was used to build an identification model based on linear discrimination analysis (LDA) using the relative reflectance values of the effective wavelengths. Other tissues were used as a prediction set to verify the reliability of the identification model. The results suggest that Vis–NIR hyperspectral images, together with the spectroscopic classification method, provide a new approach for reliable and safe diagnosis of colon cancer and could lead to advances in cancer diagnosis generally.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Bacterial Biomarkers of Marcellus Shale Activity in Pennsylvania

Unconventional oil and gas (UOG) extraction, also known as hydraulic fracturing, is becoming more prevalent with the increasing use and demand for natural gas; however, the full extent of its environmental impacts is still unknown. Here we measured physicochemical properties and bacterial community composition of sediment samples taken from twenty-eight streams within the Marcellus shale formation in northeastern Pennsylvania differentially impacted by hydraulic fracturing activities. Fourteen of the streams were classified as UOG+, and thirteen were classified as UOG- based on the presence of UOG extraction in their respective watersheds. One stream was located in a watershed that previously had UOG extraction activities but was recently abandoned. We utilized high-throughput sequencing of the 16S rRNA gene to infer differences in sediment aquatic bacterial community structure between UOG+ and UOG- streams, as well as correlate bacterial community structure to physicochemical water parameters. Although overall alpha and beta diversity differences were not observed, there were a plethora of significantly enriched operational taxonomic units (OTUs) within UOG+ and UOG- samples. Our biomarker analysis revealed many of the bacterial taxa enriched in UOG+ streams can live in saline conditions, such as Rubrobacteraceae. In addition, several bacterial taxa capable of hydrocarbon degradation were also enriched in UOG+ samples, including Oceanospirillaceae. Methanotrophic taxa, such as Methylococcales, were significantly enriched as well. Several taxa that were identified as enriched in these samples were enriched in samples taken from different streams in 2014; moreover, partial least squares discriminant analysis (PLS-DA) revealed clustering between streams from the different studies based on the presence of hydraulic fracturing along the second axis. This study revealed significant differences between bacterial assemblages within stream sediments of UOG+ and UOG- streams and identified several potential biomarkers for evaluating and monitoring the response of autochthonous bacterial communities to potential hydraulic fracturing impacts.

59 BASIC BIOLOGICAL SCIENCES↗

Arctic Tundra Vegetation Functional Types Based on Photosynthetic Physiology and Optical Properties

Climate change in tundra regions may alter vegetation species composition and ecosystem carbon balance. Remote sensing provides critical tools for monitoring these changes as optical signals provide a way to scale from plot measurements to regional patterns. Gas exchange measurements of pure patches of key vegetation functional types (lichens, mosses, and vascular plants) in sedge tundra at Barrow AK, show three significantly different values of light use efficiency (LUE) with values of 0.013+/-0.001, 0.0018+/-0.0002, and 0.0012 0.0001 mol C/mol absorbed quanta for vascular plants, mosses and lichens, respectively. Further, discriminant analysis of patch reflectance identifies five spectral bands that can separate each vegetation functional type as well as nongreen material (bare soil, standing water, and dead leaves). These results were tested along a 100 m transect where midsummer spectral reflectance and vegetation coverage were measured at one meter intervals. Area-averaged canopy LUE estimated from coverage fractions of the three functional types varied widely, even over short distances. Patch-level statistical discriminant functions applied to in situ hyperspectral reflectance successfully unmixed cover fractions of the vegetation functional types. These functions, developed from the tram data, were applied to 30 m spatial resolution Earth Observing-1 Hyperion imaging spectrometer data to examine regional variability in distribution of the vegetation functional types and from those distributions, the variability of LUE. Across the landscape, there was a fivefold variation in tundra LUE that was correlated to a spectral vegetation index developed to detect vegetation chlorophyll content.

Huemmrich, Karl F.↗

Archaeological applications of laser-induced breakdown spectroscopy: an example from the Coso Volcanic Field, California, using advanced statistical signal processing analysis

Over the past quarter century, multielement chemical analysis has become a common means for attributing the provenance of archaeological materials. The Coso Volcanic Field (CVF) in California, USA, contains at least 38 high-silica rhyolite domes, many of which contain obsidian glass that has been quarried for tools by the indigenous population for more than 12,000 years. Artifacts made from CVF obsidian are found throughout the southwestern United States and geochemical sourcing of CVF obsidian has been an important tool in understanding prehistoric Native American trading patterns. Laser-induced breakdown spectroscopy (LIBS) is a simple atomic emission spectroscopic technique that has the potential for real-time man-portable chemical analysis in the field. Because LIBS is simultaneously sensitive to all elements, a single laser shot can be used to record the broadband emission spectra, which provides a ''chemical fingerprint'' of a material. Single-shot broadband LIBS spectra were collected using a commercial benchtop LIBS system for 27 obsidian samples from major sites across the CVF and four additional sites in California and western Nevada outside of CVF. Classification of the samples was performed using partial least-squares discriminant analysis (PLSDA), a common chemometric technique suitable for performing regression on high-dimensional data. Provenance identification for the obsidian samples was evaluated for three separate labeling frameworks. The first framework consisted of a binary classification problem to distinguish CVF samples from non-CVF samples. The second approach focused on the CVF samples with labels that corresponded to the eight separate Coso sites encompassed by the 27 samples. In the third analysis, non-CVF samples were excluded, and the remaining 27 CVF samples were labeled based on groupings defined from previous major and trace element chemical studies, which reduces the number of possible classes from eight to four. Different aspects of the classifier setup considered in this study include the training/testing routine (a 27-fold leave-one-sample-out setup versus a simple split of the data into separate sets for training and evaluation), the number of latent variables used in the regression model, and whether PLSDA operating on the entire broadband LIBS spectrum is superior to that using only a selected subset of LIBS emission lines. The results point to the robustness of the PLSDA technique and suggest that LIBS analysis combined with the appropriate statistical signal processing has the potential to be a useful tool for chemical analysis of archaeological artifacts and geological specimens.

74 ATOMIC AND MOLECULAR PHYSICS↗

Arctic Tundra Vegetation Functional Types Based on Photosynthetic Physiology and Optical Properties

Non-vascular plants (lichens and mosses) are significant components of tundra landscapes and may respond to climate change differently from vascular plants affecting ecosystem carbon balance. Remote sensing provides critical tools for monitoring plant cover types, as optical signals provide a way to scale from plot measurements to regional estimates of biophysical properties, for which spatial-temporal patterns may be analyzed. Gas exchange measurements were collected for pure patches of key vegetation functional types (lichens, mosses, and vascular plants) in sedge tundra at Barrow, AK. These functional types were found to have three significantly different values of light use efficiency (LUE) with values of 0.013 plus or minus 0.0002, 0.0018 plus or minus 0.0002, and 0.0012 plus or minus 0.0001 mol C mol (exp -1) absorbed quanta for vascular plants, mosses and lichens, respectively. Discriminant analysis of the spectra reflectance of these patches identified five spectral bands that separated each of these vegetation functional types as well as nongreen material (bare soil, standing water, and dead leaves). These results were tested along a 100 m transect where midsummer spectral reflectance and vegetation coverage were measured at one meter intervals. Along the transect, area-averaged canopy LUE estimated from coverage fractions of the three functional types varied widely, even over short distances. The patch-level statistical discriminant functions applied to in situ hyperspectral reflectance data collected along the transect successfully unmixed cover fractions of the vegetation functional types. The unmixing functions, developed from the transect data, were applied to 30 m spatial resolution Earth Observing-1 Hyperion imaging spectrometer data to examine variability in distribution of the vegetation functional types for an area near Barrow, AK. Spatial variability of LUE was derived from the observed functional type distributions. Across this landscape, a fivefold variation in tundra LUE was observed. LUE calculated from the functional type cover fractions was also correlated to a spectral vegetation index developed to detect vegetation chlorophyll content. The concurrence of these alternate methods suggest that hyperspectral remote sensing can distinguish functionally distinct vegetation types and can be used to develop regional estimates of photosynthetic LUE in tundra landscapes.

VEGETATION↗

Wound outcome in combat injuries is associated with a unique set of protein biomarkers

Background: The ability to forecast whether a wound will heal after closure without further debridement(s), would provide substantial benefits to patients with severe extremity trauma. Methods: Wound effluent is a readily available material which can be collected without disturbing healthy tissue. For analysis of potential host response biomarkers, forty four serial combat wound effluent samples from 19 patients with either healing or failing traumatic- and other combat-related wounds were examined by 2-D DIGE. Spot map patterns were correlated to eventual wound outcome (healed or wound failure) and analyzed using DeCyder 7.0 and differential proteins identified via LC-MS/MS. Results: This approach identified 52 protein spots that were differentially expressed and thus represent candidate biomarkers for this clinical application. Many of these proteins are intimately involved in inflammatory and immune responses. Furthermore, discriminate analysis further refined the 52 differential protein spots to a smaller subset of which successfully differentiate between wounds that will heal and those that will fail and require further surgical intervention with greater than 83% accuracy. Conclusion: These results suggest candidates for a panel of protein biomarkers that may aid traumatic wound care prognosis and treatment. We recommend that this strategy be refined, and then externally validated, in future studies of traumatic wounds.

2-D DIGE↗

Study on Closely Related Citrus CMMs based on Chemometrics and Prediction of Components-Targets-Diseases Network by Ingenuity Pathway Analysis

As the representatives of closely related Chinese medicinal materials (CMMs) originated from Rutaceae family, Aurantii fructus immaturus (AFI), Aurantii fructus (AF), Citri reticulatae pericarpium viride (CRPV), and Citri reticulatae pericarpium (CRP) have better functions in regulating qi and promoting gastrointestinal motility. However, differences in the quality of closely related Citrus CMMs have not yet been revealed until now. Herein, this study focused on the systematic differentiation and in-depth understanding of closely related Citrus CMMs by a strategy integrating chemometrics and network pharmacology. Determined by ultra performance liquid chromatography, the content of nine flavonoids showed obvious fluctuations in the decoction pieces from different species (Citrus aurantium Linnaeus and Citrus reticulate Blanco) with decreasing levels in the samples of ripe fruits. Decoction pieces from the different species and ripening stages were well distinguished by orthogonal projection to latent structure-discriminate analysis (OPLS-DA) and cluster analysis. As a result, four active components including narirutin, naringenin, hesperidin, and 3,5,6,7,8,3′,4′-heptemthoxyflavone were filtered out by variable importance for the projection (VIP) value (VIP > 1.0), which were regarded as chemotaxonomic markers. Furthermore, a components-targets-diseases network was constructed via ingenuity pathway analysis (IPA), and the correlations were predicted between four chemotaxonomic markers, 223 targets, and three diseases including colitis, breast cancer, and colorectal cancer. The obtained results will be of great significance for identifying closely related Citrus CMMs and conduce to improving the resource utilization of CMMs.

Mu, Qixuan↗

Prediction of Response to Preoperative Chemoradiotherapy in Rectal Cancer by Multiplex Kinase Activity Profiling

Purpose: Tumor response of rectal cancer to preoperative chemoradiotherapy (CRT) varies considerably. In experimental tumor models and clinical radiotherapy, activity of particular subsets of kinase signaling pathways seems to predict radiation response. This study aimed to determine whether tumor kinase activity profiles might predict tumor response to preoperative CRT in locally advanced rectal cancer (LARC). Methods and Materials: Sixty-seven LARC patients were treated with a CRT regimen consisting of radiotherapy, fluorouracil, and, where possible, oxaliplatin. Pretreatment tumor biopsy specimens were analyzed using microarrays with kinase substrates, and the resulting substrate phosphorylation patterns were correlated with tumor response to preoperative treatment as assessed by histomorphologic tumor regression grade (TRG). A predictive model for TRG scores from phosphosubstrate signatures was obtained by partial-least-squares discriminant analysis. Prediction performance was evaluated by leave-one-out cross-validation and use of an independent test set. Results: In the patient population, 73% and 15% were scored as good responders (TRG 1-2) or intermediate responders (TRG 3), whereas 12% were assessed as poor responders (TRG 4-5). In a subset of 7 poor responders and 12 good responders, treatment outcome was correctly predicted for 95%. Application of the prediction model on the remaining patient samples resulted in correct prediction for 85%. Phosphosubstrate signatures generated by poor-responding tumors indicated high kinase activity, which was inhibited by the kinase inhibitor sunitinib, and several discriminating phosphosubstrates represented proteins derived from signaling pathways implicated in radioresistance. Conclusions: Multiplex kinase activity profiling may identify functional biomarkers predictive of tumor response to preoperative CRT in LARC.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Automated Characterization of Spent Fuel through the Multi-Isotope Process (MIP) Monitor

This research developed an algorithm for characterizing spent nuclear fuel (SNF) samples based on simulated gamma spectra. The gamma spectra for a variety of light water reactor fuels typical of those found in the United States were simulated. Fuel nuclide concentrations were simulated in ORIGEN-ARP for 1296 fuel samples with a variety of reactor designs, initial enrichments, burn ups, and cooling times. The results of the ORIGEN-ARP simulation were then input to SYNTH to simulate the gamma spectrum for each sample. These spectra were evaluated with partial least squares (PLS)-based multivariate analysis methods to characterize the fuel according to reactor type (pressurized or boiling water reactor), enrichment, burn up, and cooling time. Characterizing some of the features in series by using previously estimated features in the prediction greatly improves the performance. By first classifying the spent fuel reactor type and then using type-specific models, the prediction error for enrichment, burn up, and cooling time improved by a factor of two to four. For some features, the prediction was further improved by including additional information, such as including the predicted burn up in the estimation of cooling time. The optimal prediction flow was determined based on the simulated data. A PLS discriminate analysis model was developed which perfectly classified SNF reactor type. Burn up was predicted within 0.1% root mean squared percent error (RMSPE) and both cooling time and initial enrichment within approximately 2% RMSPE.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Fast Solution in Sparse LDA for Binary Classification

An algorithm that performs sparse linear discriminant analysis (Sparse-LDA) finds near-optimal solutions in far less time than the prior art when specialized to binary classification (of 2 classes). Sparse-LDA is a type of feature- or variable- selection problem with numerous applications in statistics, machine learning, computer vision, computational finance, operations research, and bio-informatics. Because of its combinatorial nature, feature- or variable-selection problems are NP-hard or computationally intractable in cases involving more than 30 variables or features. Therefore, one typically seeks approximate solutions by means of greedy search algorithms. The prior Sparse-LDA algorithm was a greedy algorithm that considered the best variable or feature to add/ delete to/ from its subsets in order to maximally discriminate between multiple classes of data. The present algorithm is designed for the special but prevalent case of 2-class or binary classification (e.g. 1 vs. 0, functioning vs. malfunctioning, or change versus no change). The present algorithm provides near-optimal solutions on large real-world datasets having hundreds or even thousands of variables or features (e.g. selecting the fewest wavelength bands in a hyperspectral sensor to do terrain classification) and does so in typical computation times of minutes as compared to days or weeks as taken by the prior art. Sparse LDA requires solving generalized eigenvalue problems for a large number of variable subsets (represented by the submatrices of the input within-class and between-class covariance matrices). In the general (fullrank) case, the amount of computation scales at least cubically with the number of variables and thus the size of the problems that can be solved is limited accordingly. However, in binary classification, the principal eigenvalues can be found using a special analytic formula, without resorting to costly iterative techniques. The present algorithm exploits this analytic form along with the inherent sequential nature of greedy search itself. Together this enables the use of highly-efficient partitioned-matrix-inverse techniques that result in large speedups of computation in both the forward-selection and backward-elimination stages of greedy algorithms in general.

Moghaddam, Baback↗

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↗

High-resolution time-of-flight mass spectrometry fingerprinting of metabolites from cecum and distal colon contents of rats fed resistant starch

Time-of-flight mass spectrometry along with statistical analysis was utilized to study metabolic profiles among rats fed resistant starch (RS) diets. Fischer 344 rats were fed four starch diets consisting of 55 % (w/w, dbs) starch. A control starch diet consisting of corn starch was compared against three RS diets. The RS diets were high-amylose corn starch (HA7), HA7 chemically modified with octenyl succinic anhydride, and stearic-acid-complexed HA7 starch. A subgroup received antibiotic treatment to determine if perturbations in the gut microbiome were long lasting. A second subgroup was treated with azoxymethane (AOM), a carcinogen. At the end of the 8-week study, cecal and distal colon content samples were collected from the sacrificed rats. Metabolites were extracted from cecal and distal colon samples into acetonitrile. The extracts were then analyzed on an accurate-mass time-of-flight mass spectrometer to obtain their metabolic profile. The data were analyzed using partial least-squares discriminant analysis (PLS-DA). The PLS-DA analysis utilized a training set and verification set to classify samples within diet and treatment groups. PLS-DA could reliably differentiate the diet treatments for both cecal and distal colon samples. The PLS-DA analyses of the antibiotic and no antibiotic-treated subgroups were well classified for cecal samples and modestly separated for distal colon samples. PLS-DA analysis had limited success separating distal colon samples for rats given AOM from those not treated; the cecal samples from AOM had very poor classification. Mass spectrometry profiling coupled with PLS-DA can readily classify metabolite differences among rats given RS diets.

36 MATERIALS SCIENCE↗

Mass Spectrometry and Fourier Transform Infrared Spectroscopy for Analysis of Biological Materials

Time-of-flight mass spectrometry along with statistical analysis was utilized to study metabolic profiles among rats fed resistant starch (RS) diets. Fischer 344 rats were fed four starch diets consisting of 55% (w/w, dbs) starch. A control starch diet consisting of corn starch was compared against three RS diets. The RS diets were high-amylose corn starch (HA7), HA7 chemically modified with octenyl succinic anhydride, and stearic-acid-complexed HA7 starch. A subgroup received antibiotic treatment to determine if perturbations in the gut microbiome were long lasting. A second subgroup was treated with azoxymethane (AOM), a carcinogen. At the end of the eight week study, cecal and distal-colon contents samples were collected from the sacrificed rats. Metabolites were extracted from cecal and distal colon samples into acetonitrile. The extracts were then analyzed on an accurate-mass time-of-flight mass spectrometer to obtain their metabolic profile. The data were analyzed using partial least-squares discriminant analysis (PLS-DA). The PLS-DA analysis utilized a training set and verification set to classify samples within diet and treatment groups. PLS-DA could reliably differentiate the diet treatments for both cecal and distal colon samples. The PLS-DA analyses of the antibiotic and no antibiotic treated subgroups were well classified for cecal samples and modestly separated for distal-colon samples. PLS-DA analysis had limited success separating distal colon samples for rats given AOM from those not treated; the cecal samples from AOM had very poor classification. Mass spectrometry profiling coupled with PLS-DA can readily classify metabolite differences among rats given RS diets.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Characterization of masses in digital breast tomosynthesis: Comparison of machine learning in projection views and reconstructed slices

Purpose: In digital breast tomosynthesis (DBT), quasi-three-dimensional (3D) structural information is reconstructed from a small number of 2D projection view (PV) mammograms acquired over a limited angular range. The authors developed preliminary computer-aided diagnosis (CADx) methods for classification of malignant and benign masses and compared the effectiveness of analyzing lesion characteristics in the reconstructed DBT slices and in the PVs. Methods: A data set of MLO view DBT of 99 patients containing 107 masses (56 malignant and 51 benign) was collected at the Massachusetts General Hospital with IRB approval. The DBTs were obtained with a GE prototype system which acquired 11 PVs over a 50 deg. arc. The authors reconstructed the DBTs at 1 mm slice interval using a simultaneous algebraic reconstruction technique. The region of interest (ROI) containing the mass was marked by a radiologist in the DBT volume and the corresponding ROIs on the PVs were derived based on the imaging geometry. The subsequent processes were fully automated. For classification of masses using the DBT-slice approach, the mass on each slice was segmented by an active contour model initialized with adaptive k-means clustering. A spiculation likelihood map was generated by analysis of the gradient directions around the mass margin and spiculation features were extracted from the map. The rubber band straightening transform (RBST) was applied to a band of pixels around the segmented mass boundary. The RBST image was enhanced by Sobel filtering in the horizontal and vertical directions, from which run-length statistics texture features were extracted. Morphological features including those from the normalized radial length were designed to describe the mass shape. A feature space composed of the spiculation features, texture features, and morphological features extracted from the central slice alone and seven feature spaces obtained by averaging the corresponding features from three to 19 slices centered at the central slice were compared. For classification of masses using the PV approach, a feature extraction process similar to that described above for the DBT approach was performed on the ROIs from the individual PVs. Six feature spaces obtained from the central PV alone and by averaging the corresponding features from three to 11 PVs were formed. In each feature space for either the DBT-slice or the PV approach, a linear discriminant analysis classifier with stepwise feature selection was trained and tested using a two-loop leave-one-case-out resampling procedure. Simplex optimization was used to guide feature selection automatically within the training set in each leave-one-case-out cycle. The performance of the classifiers was evaluated by the area (A{sub z}) under the receiver operating characteristic curve. Results: The test A{sub z} values from the DBT-slice approach ranged from 0.87{+-}0.03 to 0.93{+-}0.02, while those from the PV approach ranged from 0.78{+-}0.04 to 0.84{+-}0.04. The highest test A{sub z} of 0.93{+-}0.02 from the nine-DBT-slice feature space was significantly (p=0.006) better than the highest test A{sub z} of 0.84{+-}0.04 from the nine-PV feature space. Conclusion: The features of breast lesions extracted from the DBT slices consistently provided higher classification accuracy than those extracted from the PV images.

60 APPLIED LIFE SCIENCES↗

Phase I Study to Characterize the Market Potential for Non-Motorized Travel (Final Report)

The idea of livable communities suggests that people should have the option to utilize non-motorized travel (NMT), specifically walking and bicycling, to conduct their daily tasks. Forecasting personal travel by walk and bike is necessary as part of regional transportation planning, and requires fine detail not only about individual travel, but also on transportation and neighborhood infrastructure. In an attempt to characterize the 'market' potential for NMT, the Office of Planning, Federal Highway Administration (FHWA) funded the Center for Transportation Analysis (CTA) of the Oak Ridge National Laboratory (ORNL) to conduct a study. The objectives of this effort were to identify factors that influence communities to walk and bike and to examine why, or why not, travelers walk and bike in their communities. This study relied on information collected under the 2009 National Household Travel Survey (NHTS) as the major source of data, and was supplemented with data from the American Community Survey (ACS), educational survey, health, employment, and others. Initial statistical screening methods were applied to sort through over 400 potential predictor variables, and examined with various measures (e.g., walk trip per person, walk mileage per person, bike trip per person, bike mileage per person) as the dependent variables. The best geographic level of detail used in the modeling for this study was determined to be the Census block group level for walking and Census tract level for biking. The need for additional supplemental private data (i.e., Walk Scores and Nielsen employment data), and geospatial information that reflects land use and physical environments, became evident after an examination of findings from the initial screening models. To be feasible, in terms of costs and time, the geographic scale of the study region was scaled down to nine selected NHTS add-on regions. These regions were chosen based on various criteria including transit availability, population size, and a mix of geographic locations across the nation. Given the similarities in modeling results from walk trips and walk mileages, additional modeling efforts conducted under the later part of this study were focused on walk trips per person. Bike models were limited only with the stepwise logistic models using Census tracts in the selected regions. Due to NHTS sampling limitations, only about 12% of these tracts have bike trips recorded from NHTS sampled households. The modeling with NHTS bike data proved to be more challenging and time consuming than what was anticipated. Along with the late arrival of Nielsen employment data, the project team had to limit the modeling effort to focus on walking. Therefore, the final modeling and discriminant analysis was conducted only for walking trips.

54 ENVIRONMENTAL SCIENCES↗