Method For Cross-validation Research Articles

BrGDGT-based seasonal paleotemperature reconstruction for the last 15 000 years from a shallow lake on the eastern Tibetan Plateau

Abstract. Understanding Holocene temperature changes is vital for resolving discrepancies between proxy reconstructions and climate models. The intricate temperature variations across the Tibetan Plateau (TP) add complexity to studying continental climate change during this period. Discrepancies between model-based and proxy-based reconstructions might stem from seasonal biases and environmental uncertainties in the proxies. Employing multiple proxies from a single sediment core for quantitative temperature reconstructions offers an effective method for cross-validation in terrestrial environments. Here, we present an ice-free-season temperature record for the past 15 kyr from a shallow, freshwater lake on the eastern TP, based on brGDGTs (branched glycerol dialkyl glycerol tetraethers). This record shows that the Holocene Thermal Maximum lags the pollen-based July temperature recorded in the same sediment core. We conclude that the mismatch between the brGDGT-based and pollen-based temperatures is primarily the result of seasonal variations in solar irradiance. The overall pattern of temperature changes is supported by other summer temperature records, and the Younger Dryas cold event and the Bølling–Allerød warm period are also detected. A generally warm period occurred during 8–3.5 ka, followed by a cooling trend in the late Holocene. Our findings have implications for understanding the seasonal signal of brGDGTs in shallow lakes and provide critical data for confirming the occurrence of seasonal biases in different proxies from high-elevation lakes. To further investigate the significance of the brGDGTs and temperature patterns on the TP, we examined existing brGDGT-based Holocene temperature records, which interpret these compounds as indicators of mean annual or growing season temperatures. The existing and available temperature records show complicated patterns of variation, some with general warming trends throughout the Holocene, some with cooling trends, and some with a warm middle Holocene. We analyzed the possible reasons for the diverse brGDGTs records on the TP and emphasize the importance of considering lake conditions and modern investigations of brGDGTs in lacustrine systems when using brGDGTs to reconstruct paleoenvironmental conditions.

Rank-ordering of known enzymes as starting points for re-engineering novel substrate activity using a convolutional neural network

Retro-biosynthetic approaches have made significant advances in predicting synthesis routes of target biofuel, bio-renewable or bio-active molecules. The use of only cataloged enzymatic activities limits the discovery of new production routes. Recent retro-biosynthetic algorithms increasingly use novel conversions that require altering the substrate or cofactor specificities of existing enzymes while connecting pathways leading to a target metabolite. However, identifying and re-engineering enzymes for desired novel conversions are currently the bottlenecks in implementing such designed pathways. Herein, we present EnzRank, a convolutional neural network (CNN) based approach, to rank-order existing enzymes in terms of their suitability to undergo successful protein engineering through directed evolution or de novo design towards a desired specific substrate activity. We train the CNN model on 11,800 known active enzyme-substrate pairs from the BRENDA database as positive samples and data generated by scrambling these pairs as negative samples using substrate dissimilarity between an enzyme's native substrate and all other molecules present in the dataset using Tanimoto similarity score. EnzRank achieves an average recovery rate of 80.72% and 73.08% for positive and negative pairs on test data after using a 10-fold holdout method for training and cross-validation. We further developed a web-based user interface (available at https://huggingface.co/spaces/vuu10/EnzRank) to predict enzyme-substrate activity using SMILES strings of substrates and enzyme sequence as input to allow convenient and easy-to-use access to EnzRank. In summary, this effort can aid de novo pathway design tools to prioritize starting enzyme re-engineering candidates for novel reactions as well as in predicting the potential secondary activity of enzymes in cell metabolism.

Reasons, willingness, and intention to be a foster family: A community-sample study

Research into the reasons that could lead community samples to become foster families is sparse, as most studies focus on the reasons identified by licensed or prospective foster families. The present study aims: (1) to assess the validity and reliability of the Reasons for Fostering Inventory adapted for a community sample (Portuguese version); and (2) to test the role of family factors and the different reasons for fostering as potential predictors of willingness and behavioral intention to foster. The reliability and validity of the inventory were assessed using a convenience sample of Portuguese adults (n = 441), 84 % female, aged 26–74. A three-factorial structure (self-oriented reasons, child-centered reasons, and family-related reasons) composed of 22 items was identified. The factorial structure was tested with a holdout randomization method for cross-validation. All factors were reliable, with internal consistency levels ranging from 0.85 and 0.88. Being female, younger, and scoring lower on family-related reasons and greater on child-centered reasons showed positive associations with both willingness and intention to foster. Lower scores on self-oriented reasons were associated with higher willingness to foster; while having parental experience, lower education, and greater income were associated with a higher intention to foster. This study contributes with additional psychometric evidence for this scale for use with community samples. It also provides new insight into how individual resources may be linked with willingness and intention to foster in a community sample.

Deep learning models for RNA secondary structure prediction (probably) do not generalize across families.

MotivationThe secondary structure of RNA is of importance to its function. Over the last few years, several papers attempted to use machine learning to improve de novo RNA secondary structure prediction. Many of these papers report impressive results for intra-family predictions but seldom address the much more difficult (and practical) inter-family problem.ResultsWe demonstrate that it is nearly trivial with convolutional neural networks to generate pseudo-free energy changes, modelled after structure mapping data that improve the accuracy of structure prediction for intra-family cases. We propose a more rigorous method for inter-family cross-validation that can be used to assess the performance of learning-based models. Using this method, we further demonstrate that intra-family performance is insufficient proof of generalization despite the widespread assumption in the literature and provide strong evidence that many existing learning-based models have not generalized inter-family.Availability and implementationSource code and data are available at https://github.com/marcellszi/dl-rna.Supplementary information Supplementary data are available at Bioinformatics online.

Relaxation Degree Analysis Using Frontal Electroencephalogram Under Virtual Reality Relaxation Scenes.

Increasing social pressure enhances the psychological burden on individuals, and the severity of depression can no longer be ignored. The characteristics of high immersion and interactivity enhance virtual reality (VR) application in psychological therapy. Many studies have verified the effectiveness of VR relaxation therapy, although a few have performed a quantitative study on relaxation state (R-state). To confirm the effectiveness of VR relaxation and quantitatively assess relaxation, this study confirmed the effectiveness of the VR sightseeing relaxation scenes using subjective emotion scale and objective electroencephalogram (EEG) data from college students. Moreover, some EEG features with significant consistent differences after they watched the VR scenes were detected including the energy ratio of the alpha wave, gamma wave, and differential asymmetry. An R-state regression model was then built using the model stacking method for optimization, of which random forest regression, AdaBoost, gradient boosting (GB), and light GB were adopted as the first level, while linear regression and support vector machine were applied at the second level. The leave-one-subject-out method for cross-validation was used to evaluate the results, where the mean accuracy of the framework achieved 81.46%. The significantly changed features and the R-state model with over 80% accuracy have laid a foundation for further research on relaxation interaction systems. Moreover, the VR relaxation therapy was applied to the clinical treatment of patients with depression and achieved preliminary good results, which might provide a possible method for non-drug treatment of patients with depression.

Integrated Herfindahl–Hirschman Index, Compromise Programming, and ε-Constraint Method For Multicriteria Performance-Based Transportation Budget Allocation

This paper introduces a new method for multicriteria performance-based transportation budget allocation. First, the Herfindahl–Hirschman Index (HHI) that has long been used in economics for measuring the level of monopolization in a marketplace is employed to derive the relative weights of multiple non-commensurable transportation performance criteria. Next, a compromise programming (CP) model is formulated to help transform the multicriteria optimization formulation for transportation budget allocation to a single-objective optimization model solvable by minimizing the Chebyshev distance to the ideal levels of performance targets associated with individual performance criteria. Finally, the ε-constraint trade-off analysis method is incorporated into the combined use of HHI and CP framework to iteratively derive the optimal decision outcome. Six-year data on candidate investment projects proposed for a U.S. state-maintained rural Interstate highway system along with data details of the available budget is used in a computational study for method application. Comparative analysis of decision outcomes is conducted with the STEP method for cross validation. It has revealed that the proposed HHI-CP ε-constraint method outperforms the existing method and can be adopted by state and local transportation agencies to carry out effective budget allocation.

Composite Machine Learning Algorithm for Material Sourcing.

This study developed a composite machine learning algorithm for attribution of materials of forensic interest (like ammonium nitrate) to original sources. k-nearest neighbor and random forest models were used for source elimination and classification, respectively, in a two-step, composite algorithm based on particle color, size/shape, and trace element concentration features. Novel approaches for simulation to supplement within-source reference features based on empirically measured multi-lot analyses, an improved hold-one-lot-out method for cross-validation, an assessment of the likelihood of the presence of a reference sample, fusion of the source probabilities from the respective classification models, and the calculation of metrics for assessing ensemble sourcing performance are described. Excellent sourcing predictions were obtained; the sourcing algorithm identified the correct source as the top choice 89% of the time, and the correct source was identified to be an average of 2.7 times more likely than the most likely incorrect source.

High-performance prediction of macauba fruit biomass for agricultural and industrial purposes using Artificial Neural Networks

Biomass estimation plays of crucial role in agriculture and agro-based industries. The macauba, Acrocomia aculeata (Jacq.) Lood., ex Mart., is a palm species that has been a focal point for research and development of an alternative biomass-bioenergy crop for the tropics. The macauba fruit components (exocarp, mesocarp, endocarp and seed/kernel) present different constitutional characteristics and their biomass determination, by traditional methods, is labor-consuming. Therefore, the validation of procedures that can streamline this process is relevant, since it can reduce costs and time for both breeding programs and industries. This study tested the efficacy of Artificial Neural Networks (ANN) on biomass prediction of the macauba fruit components by comparing it to the multiple linear regression method. The data used came from fruits collected in 18 localities, distributed throughout the state of Minas Gerais, Brazil. According to their provenance, the matrices were clustered into two groups with the k-means method for posterior ANN cross-validation. Each group was interchangeably used for both training and validation purposes. The ANN was more efficient than multivariate linear model in the predictions of dry weight of the fruit́s four components and oil content of the mesocarp and seed. As for variables related to dry weight, ANN reached 98% predictive accuracy (i.e., 98% accuracy of the value predicted by the network), and for variables related to oil contents, accuracy was around 90%. Additionally, non-invasive measurements of the fruit (i.e., low-cost and low-time measurement variables) were adequate enough to predict most of the variables of interest. These results show the ANN's prediction potential, saving time and efforts for the consolidation of macauba as a crop.

Fast cross validation for regularized extreme learning machine

A method for fast l-fold cross validation is proposed for the regularized extreme learning machine (RELM). The computational time of fast l-fold cross validation increases as the fold number decreases, which is opposite to that of naive l-fold cross validation. As opposed to naive l-fold cross validation, fast l-fold cross validation takes the advantage in terms of computational time, especially for the large fold number such as l > 20. To corroborate the efficacy and feasibility of fast l-fold cross validation, experiments on five benchmark regression data sets are evaluated.

Abstract 2376: MutationValidator: A computational method for variant cross-validation in next-generation sequencing data

Abstract Background: Massively parallel next-generation sequencing (NGS) technologies already enable comprehensive genome-wide detection of germline and somatic variants in thousands of individuals, and the pace of data production keeps accelerating. However, with increased sequencing techniques and variant calling algorithms, proper validation and determination of ground truth is essential for benchmarking, comparisons, and confirmation of novel discoveries. Datasets generated on different sequencing platforms, derived from distinct biological samples and/or using different library construction protocols can provide at least partially independent sources for crosschecking variants. While methods for validation exist to date, they fail to account for the inherent variability in the datasets, and thereby propose biased estimates for the validation rates. Results: Here we present a method for variant cross-validation and present software implementation (written in Matlab) that generates validation metrics and diagnostics to characterize a mutation as true somatic, germline, or artifactual. Our method adjusts for the power to observe variant-supporting evidence in the validation dataset, which allows for accurate assessment of validation rates across wide ranges of sequencing coverages and variant allelic fractions. When this method is applied to a test cohort of 172 patients from The Cancer Genome Atlas Lung Adenocarcenoma dataset, we observe a SNP and Indel validation rate of 95% and 96% respectively with a 99% power for detection. A previously implemented naïve method of >8% supporting alleles in the tumor and <3% alternate allele in the normal for a validated event would have yielded rates of 72% and 95% for SNPs and Indels, values we claim to be underestimations of the true rates as they misclassify underpowered sites as a false positives. Conclusions: Our proposed method, MutationValidator, enables confident classification of mutations using orthogonal next-generation sequencing technologies for validation as it takes into consideration the variability in and across the datasets. Citation Format: Mara Rosenberg, Gad Getz, Adam Kiezun, Andrey Sivachenko. MutationValidator: A computational method for variant cross-validation in next-generation sequencing data. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 2376. doi:10.1158/1538-7445.AM2014-2376

Measures of angular spread and entropy for the detection of architectural distortion in prior mammograms

Architectural distortion is an important sign of early breast cancer. We present methods for computer-aided detection of architectural distortion in mammograms acquired prior to the diagnosis of breast cancer in the interval between scheduled screening sessions. Potential sites of architectural distortion were detected using node maps obtained through the application of a bank of Gabor filters and linear phase portrait modeling. A total of 4,224 regions of interest (ROIs) were automatically obtained from 106 prior mammograms of 56 interval-cancer cases, including 301 true-positive ROIs, and from 52 mammograms of 13 normal cases. Each ROI was represented by three types of entropy measures of angular histograms composed with the Gabor magnitude response, angle, coherence, orientation strength, and the angular spread of power in the Fourier spectrum, including Shannon's entropy, Tsallis entropy for nonextensive systems, and Rényi entropy for extensive systems. Using the entropy measures with stepwise logistic regression and the leave-one-patient-out method for feature selection and cross-validation, an artificial neural network resulted in an area under the receiver operating characteristic curve of 0.75. Free-response receiver operating characteristics indicated a sensitivity of 0.80at 5.2 false positives (FPs) per patient. The proposed methods can detect architectural distortion in prior mammograms taken 15months (on the average) before clinical diagnosis of breast cancer, with a high sensitivity and a moderate number of FPs per patient. The results are promising and may be improved with additional features to characterize subtle abnormalities and larger databases including prior mammograms.

Pretreatment metabotype as a predictor of response to sertraline or placebo in depressed outpatients: a proof of concept

The purpose of this study was to determine whether the baseline metabolic profile (that is, metabotype) of a patient with major depressive disorder (MDD) would define how an individual will respond to treatment. Outpatients with MDD were randomly assigned to sertraline (up to 150 mg per day) (N=43) or placebo (N=46) in a double-blind 4-week trial. Baseline serum samples were profiled using the liquid chromatography electrochemical array; the output was digitized to create a ‘digital map' of the entire measurable response for a particular sample. Response was defined as ⩾50% reduction baseline to week 4 in the 17-item Hamilton Rating Scale for Depression total score. Models were built using the one-out method for cross-validation. Multivariate analyses showed that metabolic profiles partially separated responders and non-responders to sertraline or to placebo. For the sertraline models, the overall correct classification rate was 81% whereas it was 72% for the placebo models. Several pathways were implicated in separation of responders and non-responders on sertraline and on placebo including phenylalanine, tryptophan, purine and tocopherol. Dihydroxyphenylacetic acid, tocopherols and serotonin were common metabolites in separating responders and non-responders to both drug and placebo. Pretreatment metabotypes may predict which depressed patients will respond to acute treatment (4 weeks) with sertraline or placebo. Some pathways were informative for both treatments whereas other pathways were unique in predicting response to either sertraline or placebo. Metabolomics may inform the biochemical basis for the early efficacy of sertraline.

Correcting evaluation bias of relational classifiers with network cross validation

Recently, a number of modeling techniques have been developed for data mining and machine learning in relational and network domains where the instances are not independent and identically distributed (i.i.d.). These methods specifically exploit the statistical dependencies among instances in order to improve classification accuracy. However, there has been little focus on how these same dependencies affect our ability to draw accurate conclusions about the performance of the models. More specifically, the complex link structure and attribute dependencies in relational data violate the assumptions of many conventional statistical tests and make it difficult to use these tests to assess the models in an unbiased manner. In this work, we examine the task of within-network classification and the question of whether two algorithms will learn models that will result in significantly different levels of performance. We show that the commonly used form of evaluation (paired t-test on overlapping network samples) can result in an unacceptable level of Type I error. Furthermore, we show that Type I error increases as (1) the correlation among instances increases and (2) the size of the evaluation set increases (i.e., the proportion of labeled nodes in the network decreases). We propose a method for network cross-validation that combined with paired t-tests produces more acceptable levels of Type I error while still providing reasonable levels of statistical power (i.e., 1−Type II error).

Reannotation of protein‐coding genes based on an improved graphical representation of DNA sequence

Over annotation of protein coding genes is common phenomenon in microbial genomes, the genome of Amsacta moorei entomopoxvirus (AmEPV) is a typical case, because more than 63% of its annotated ORFs are hypothetical. In this article, we propose an improved graphical representation titled I-TN (improved curve based on trinucleotides) curve, which allows direct inspection of composition and distribution of codons and asymmetric gene structure. This improved graphical representation can also provide convenient tools for genome analysis. From this presentation, 18 variables are exploited as numerical descriptors to represent the specific features of protein coding genes quantitatively, with which we reannotate the protein coding genes in several viral genomes. Using the parameters trained on the experimentally validated genes, all of the 30 experimentally validated genes and 63 putative genes in AmEPV genome are recognized correctly as protein coding, the accuracies of the present method for self-test and cross-validation are 100%, respectively. Twenty-eight annotated hypothetical genes are predicted as noncoding, and then the number of reannotated protein coding genes in AmEPV should be 266 instead of 294 reported in the original annotations. Extending the present method trained in AmEPV to other entomopoxvirus genomes directly, such as Melanoplus sanguinipes entomopoxvirus (MsEPV), all of the 123 annotated function-known and putative genes are recognized correctly as protein coding, and 17 hypothetical genes are recognized as noncoding. The present method could also be extended to other genomes with or without adaptation of training sets with high accuracy.

Efficient Leave-m-out Cross-Validation of Support Vector Regression by Generalizing Decremental Algorithm

We propose a computationally efficient method for cross-validation of the Support Vector Regression (SVR) by generalizing the decremental algorithm of SVR. Incremental and decremental algorithm of Support Vector Machines (SVM) 2, 8, 9) efficiently update the trained SVM model when a single data point is added to or removed from the training set. The computational cost of leave-one-out cross-validation can be reduced using the decremental algorithm. However, when we perform leave-m-out cross-validation (m > 1), we have to repeatedly apply the decremental algorithm for each data point. In this paper, we extend the decremental algorithm of SVR8, 9) in such a way that several data points can be removed more efficiently. Experimental results indicate that the proposed approach can reduce the computational cost. In particular, we observed that the number of breakpoints, which is the main computational cost of the involved path-following, were reduced from \({\mathcal O}(m)\) to \({\mathcal O}(\sqrt{m})\).

In vivo real time monitoring of vasoconstriction and vasodilation by a combined diffuse reflectance spectroscopy and Doppler optical coherence tomography approach

A combined diffuse reflectance (DR) spectroscopy and doppler optical coherence tomography (DOCT) approach may offer a powerful means for assessment of tissue function, and potentially provide a way for earlier cancer detection through non-invasive local blood supply measurements. The goal of the study was to compare a DR-derived blood-content-related index to a measure of local blood supply flow as furnished by DOCT during manipulations with blood circulation (vasoconstriction and vasodilation), investigate similarities and differences, complementarity of techniques, and then applying these results to the underlying biology. Simultaneous DR-DOCT measurements of local blood supply were conducted during drug and mechanically-induced vasoconstriction and vasodilatation on an externalized intact rat gut in vivo. A simple heuristic metric, termed Blood Supply Index was derived from the spectroscopic DR data. This metric variance due to mechanical and pharmacological manipulation of the local blood supply was recorded, and compared with that of two DOCT-derived metrics, namely normalized blood velocity (v(rel)) and blood vessel diameter (D). During vasoconstriction and vasodilatation, the local blood response was successfully visualized by both DOCT and DR metrics and a reproducible correlation was found between these two measurements. A combined DR-DOCT approach may evolve into a technologically-viable method for cross-validation of the derived haemodynamic metrics, yielding a more reliable functional tissue assessment tool for accurate cancer diagnosis and staging.

Highly Efficient Classification and Identification of Human Pathogenic Bacteria by MALDI-TOF MS

Accurate and rapid identification of pathogenic microorganisms is of critical importance in disease treatment and public health. Conventional work flows are time-consuming, and procedures are multifaceted. MS can be an alternative but is limited by low efficiency for amino acid sequencing as well as low reproducibility for spectrum fingerprinting. We systematically analyzed the feasibility of applying MS for rapid and accurate bacterial identification. Directly applying bacterial colonies without further protein extraction to MALDI-TOF MS analysis revealed rich peak contents and high reproducibility. The MS spectra derived from 57 isolates comprising six human pathogenic bacterial species were analyzed using both unsupervised hierarchical clustering and supervised model construction via the Genetic Algorithm. Hierarchical clustering analysis categorized the spectra into six groups precisely corresponding to the six bacterial species. Precise classification was also maintained in an independently prepared set of bacteria even when the numbers of m/z values were reduced to six. In parallel, classification models were constructed via Genetic Algorithm analysis. A model containing 18 m/z values accurately classified independently prepared bacteria and identified those species originally not used for model construction. Moreover bacteria fewer than 10(4) cells and different species in bacterial mixtures were identified using the classification model approach. In conclusion, the application of MALDI-TOF MS in combination with a suitable model construction provides a highly accurate method for bacterial classification and identification. The approach can identify bacteria with low abundance even in mixed flora, suggesting that a rapid and accurate bacterial identification using MS techniques even before culture can be attained in the near future.

Re-prediction of protein-coding genes in the genome of Amsacta moorei entomopoxvirus

Using the Z curve method, the protein-coding genes in AmEPV genome are re-predicted. On the basis of the parameters trained on the experimentally validated genes, all of the 30 experimentally validated genes and 67 putative genes are predicted correctly as coding genes. The sensitivities of the present method for self-test and cross-validation are all 100% based on these test sets. Thirty-eight annotated conserved and hypothetical genes are predicted as non-coding ORFs. The number of re-predicted protein-coding genes in AmEPV is 256. It is significantly less than the number 294 reported in the original annotation. After extending the present method trained in AeEPV genome to the other entomopoxvirus genome, it is found that 116 of the 123 known and putative genes are predicted correctly as coding. Six of the seven falsely missed genes are less than 300 bp. The present method could be extended to other poxvirus genomes with or without adaptation of training sets.

Fast cross-validation algorithms for least squares support vector machine and kernel ridge regression

Given n training examples, the training of a least squares support vector machine (LS-SVM) or kernel ridge regression (KRR) corresponds to solving a linear system of dimension n. In cross-validating LS-SVM or KRR, the training examples are split into two distinct subsets for a number of times ( l ) wherein a subset of m examples are used for validation and the other subset of ( n - m ) examples are used for training the classifier. In this case l linear systems of dimension ( n - m ) need to be solved. We propose a novel method for cross-validation (CV) of LS-SVM or KRR in which instead of solving l linear systems of dimension ( n - m ) , we compute the inverse of an n dimensional square matrix and solve l linear systems of dimension m, thereby reducing the complexity when l is large and/or m is small. Typical multi-fold, leave-one-out cross-validation (LOO-CV) and leave-many-out cross-validations are considered. For five-fold CV used in practice with five repetitions over randomly drawn slices, the proposed algorithm is approximately four times as efficient as the naive implementation. For large data sets, we propose to evaluate the CV approximately by applying the well-known incomplete Cholesky decomposition technique and the complexity of these approximate algorithms will scale linearly on the data size if the rank of the associated kernel matrix is much smaller than n . Simulations are provided to demonstrate the performance of LS-SVM and the efficiency of the proposed algorithm with comparisons to the naive and some existent implementations of multi-fold and LOO-CV.

P2-222: Classification of normal elderly, MCI, and mild AD using neuropsychological tests and self-reported memory function

Like most diseases, the most effective management in AD, and the ultimate hope for cure, will depend on early recognition and intervention. The amnestic form of mild cognitive impairment (MCI) is a frequent precursor of AD and may be the earliest clinical stage of AD. Using magnetoencephalography (MEG) in a simple 45–sec long fixation task, we demonstrated reliable changes in the strength of synchronous positive (Lewis et al., Soc. Neurosci. Abstr. 542: 6, 2005) and negative interactions (McCarten et al., Soc. Neurosci. Abstr. 542: 7, 2005) between neuronal ensembles in normal elderly (NE), amnestic MCI and mild AD subjects, and provided support for the concept that MCI represents early AD. In the present analysis, we focused on neuropsychological tests and self–reporting memory function to find out whether a set of such variables can yield correct classification of subjects to these diagnostic categories. Subjects were in good general health (NE; N=4; Age± SD =76.67±5.03; MMSE±SD= 28.50±1.29), had amnestic MCI (N=4; 75.0±7.87; 26.50±1.29) or mild AD (N=4; 79.0±0.82; 25.50±1.00) based on a comprehensive history, physical and neuropsychological evaluation. All subjects were tested on a battery of neuropsychological tests (yielding 42 variables), and also completed a 64 item self–report memory questionnaire before and after neuropsychological testing. A linear discriminant classification analysis was performed using the robust leave–one–out method for cross validation. We found that a combination of 7 specific neuropsychological and self–report indices (Digit Span forward, list recall, story recall, and 4 self–report ratings of the seriousness of one's memory impairment) was able to correctly classify all subjects even using the leave–one–out method. In addition, a different set of 6 self–reported memory functioning items, on their own, were able to correctly classify 92% of cases. We are currently assessing the power of MEG–derived synchronous dynamic networks (Langheim et al., PNAS USA 103: 455, 2006) to correctly classify NE, MCI, and AD subjects with highly promising results. These findings demonstrate that brief and non–invasive tests can be used successfully to classify subjects into one of these diagnostic categories.