Interpretable gene expression classifier with an accurate and compact fuzzy rule base for microarray data analysis

An accurate classifier with linguistic interpretability using a small number of relevant genes is beneficial to microarray data analysis and development of inexpensive diagnostic tests. Several frequently used techniques for designing classifiers of microarray data, such as support vector machine, neural networks, k-nearest neighbor, and logistic regression model, suffer from low interpretabilities. This paper proposes an interpretable gene expression classifier (named iGEC) with an accurate and compact fuzzy rule base for microarray data analysis. The design of iGEC has three objectives to be simultaneously optimized: maximal classification accuracy, minimal number of rules, and minimal number of used genes. An "intelligent" genetic algorithm IGA is used to efficiently solve the design problem with a large number of tuning parameters. The performance of iGEC is evaluated using eight commonly-used data sets. It is shown that iGEC has an accurate, concise, and interpretable rule base (1.1 rules per class) on average in terms of test classification accuracy (87.9%), rule number (3.9), and used gene number (5.0). Moreover, iGEC not only has better performance than the existing fuzzy rule-based classifier in terms of the above-mentioned objectives, but also is more accurate than some existing non-rule-based classifiers.     

基于表面肌电信号的人体动作识别算法研究进展

表面肌电信号(Surface Electromyography,sEMG)是通过相应肌群表面的传感器记录下来的一维时间序列非平稳生物电信号,不但反映了神经肌肉系统活动,对于反映相应动作肢体活动信息同样重要。而模式识别是肌电应用领域的基础和关键。为了在应用基于表面肌电信号模式识别中选取合适算法,本文拟对基于表面肌电信号的人体动作识别算法进行回顾分析,主要包括模糊模式识别算法、线性判别分析算法、人工神经网络算法和支持向量机算法。模糊模式识别能自适应提取模糊规则,对初始化规则不敏感,适合处理s EMG这样具有严格不重复的生物电信号;线性判别分析对数据进行降维,计算简单,但不适合大数据;人工神经网络可以同时描述训练样本输入输出的线性关系和非线性映射关系,可以解决复杂的分类问题,学习能力强;支持向量机处理小样本、非线性的高维数据优势明显,计算速度快。比较各方法的优缺点,为今后处理此类问题模式识别算法选取提供了参考和依据。     

‘True’ null allele detection in microsatellite loci: a comparison of methods,assessment of difficulties and survey of possible improvements

Null alleles are alleles that for various reasons fail to amplify in a PCR assay. The presence of null alleles in microsatellite data is known to bias the genetic parameter estimates. Thus, efficient detection of null alleles is crucial, but the methods available for indirect null allele detection return inconsistent results. Here, our aim was to compare different methods for null allele detection, to explain their respective performance and to provide improvements. We applied several approaches to identify the ‘true’ null alleles based on the predictions made by five different methods, used either individually or in combination. First, we introduced simulated ‘true’ null alleles into 240 population data sets and applied the methods to measure their success in detecting the simulated null alleles. The single best‐performing method was ML‐NullFreq_frequency. Furthermore, we applied different noise reduction approaches to improve the results. For instance, by combining the results of several methods, we obtained more reliable results than using a single one. Rule‐based classification was applied to identify population properties linked to the false discovery rate. Rules obtained from the classifier described which population genetic estimates and loci characteristics were linked to the success of each method. We have shown that by simulating ‘true’ null alleles into a population data set, we may define a null allele frequency threshold, related to a desired true or false discovery rate. Moreover, using such simulated data sets, the expected null allele homozygote frequency may be estimated independently of the equilibrium state of the population.     

Moist deciduous forest identification using temporal MODIS data — A comparative study using fuzzy based classifiers

The two soft fuzzy based classifiers, Possibilistic c-Means (PCM) approach and Noise Clustering (NC) were compared for the Moist Deciduous Forest (MDF) identification from MODIS temporal data. Seven date temporal MODIS data were used to identify MDF and temporal Advanced Wide Field Sensor (AWiFS) data was used as reference data for testing. Simple Ratio (SR), Normalized Difference Vegetation Index (NDVI), Soil Adjusted Vegetation Index (SAVI) and Enhanced Vegetation Index 2 (EVI2) were used to generate the temporal spectral index datasets for both the MODIS and AWiFS. The parameter weighting exponent m for PCM and resolution parameter δ for NC were optimized. Results show that the optimized value of m for MDF is 2.1, while δ value is 3.6 × 10⁴ for temporal MODIS data. For assessment of the accuracy AWiFS datasets were also optimized using entropy approach. The optimized dataset of AWiFS was then used for accuracy assessment of the soft classified outputs from MODIS using Fuzzy ERror Matrix (FERM). It was found from this study that, for PCM classifier the highest fuzzy overall accuracy of 97.44% was obtained using the SAVI for the temporal dataset ‘Five’ consisting to one scene of ‘Full greenness’, three scenes in ‘Intermediate frequency stage of Onset of Greenness (OG) and End of Senescence (ES) activity’ and the last image pertaining corresponds to the ‘Maximum frequency stage of OG activity’ as per phenology of MDF. Similarly, for NC classifier the highest fuzzy overall accuracy of 95.19% was obtained for the EVI2 with temporal dataset ‘Five’ consisting with two scene of ‘Full greenness’, two scenes in ‘Intermediate frequency stage of OG and ES activity’ and the last one corresponds to the ‘Maximum frequency stage of OG activity’as per phenology of MDF.     

Overlapped Partitioning for Ensemble Classifiers of P300-Based Brain-Computer Interfaces

A P300-based brain-computer interface (BCI) enables a wide range of people to control devices that improve their quality of life. Ensemble classifiers with naive partitioning were recently applied to the P300-based BCI and these classification performances were assessed. However, they were usually trained on a large amount of training data (e.g., 15300). In this study, we evaluated ensemble linear discriminant analysis (LDA) classifiers with a newly proposed overlapped partitioning method using 900 training data. In addition, the classification performances of the ensemble classifier with naive partitioning and a single LDA classifier were compared. One of three conditions for dimension reduction was applied: the stepwise method, principal component analysis (PCA), or none. The results show that an ensemble stepwise LDA (SWLDA) classifier with overlapped partitioning achieved a better performance than the commonly used single SWLDA classifier and an ensemble SWLDA classifier with naive partitioning. This result implies that the performance of the SWLDA is improved by overlapped partitioning and the ensemble classifier with overlapped partitioning requires less training data than that with naive partitioning. This study contributes towards reducing the required amount of training data and achieving better classification performance.     

Fuzzy knowledge- and data-based models of damage to reeds by grazing of Greylag Geese

This paper describes a fuzzy and neuro-fuzzy approach to modelling feeding intensity of Greylag Geese on reed. As a consequence of the presence of some non-measurable or random factors and the heterogeneity of reed and goose behaviour, the relationships between the model variables are often not well known and the data collected have a high degree of uncertainty. A fuzzy approach was selected which can be applied with vague knowledge and data of high uncertainty. Fuzzy logic can be used to handle inexact reasoning in knowledge-based models with fuzzy rules and fuzzy sets to handle uncertainty in data. The neural network technique was applied to develop the fuzzy data-based models. For training, several dataset combinations of three lakes in North Germany were used. The generalisation capability of these models was checked for other lakes. The performance of these models was compared with the results of the fuzzy knowledge-based model developed in the next step. The knowledge base of this model contains the Mamdani-type rules formulated by a domain expert. All models were implemented using the Fuzzy Logic Toolbox of MATLAB^®.     

Comparative phylogeography in marginal seas of the northwestern Pacific

The maturation of marine phylogeography depends on integration of comparative information across different regions globally. The northwestern Pacific, characterized by unique tectonic setting, however, is still underrepresented. This study seeks to highlight its phylogeographical history based on the available population data, focusing on three seas: the East China Sea (ECS), the South China Sea (SCS) and the Sea of Japan (SOJ). We first conducted a literature survey to evaluate current research efforts and then reanalysed the population structure, historical demography and genealogy for two selections of studies (namely ‘the ECS category’ and ‘the multiple‐sea category’) to elucidate the evolutionary processes within and across the seas, respectively. For the ECS category, the meta‐analyses revealed most studies displayed a shallow phylogeny, indicating a single origin from the sea. Significant population structure was commonplace, particularly in molluck and crustacean studies, with proportions of 89% and 80%, respectively. Nearly all studies selected showed signals of population expansion: the times estimated were closely linked to a period of ~120–140 Kya rather than the last glacial maximum. For the latter category, divergent intraspecific lineages appeared among seas and overlapped in the adjacent regions, a pattern implying each sea had served as an independent refugium during glaciations. The genetic splits, however, were estimated to arise from separate events dating from late Miocene to middle Pleistocene. As phylogeography is still in its infancy in the region, more effort is needed to test and complement the general rules abstracted here. Finally, challenges and prospects were discussed to accelerate further research.     

Noise affects porpoise click detections – the magnitude of the effect depends on logger type and detection filter settings

Automatic click detectors and full-bandwidth sound recorders are widely used in passive acoustic monitoring of small cetaceans. Detection of these signals depends on a variety of factors, including signal to noise ratio. Passive acoustic monitoring is often used to study impact of underwater noise on small cetaceans, but as detection probability is affected by changes in signal to noise ratio, variable noise levels may affect conclusions drawn from these experiments. Therefore, we examine how different detectors and filters perform in varying ocean noise conditions. C-PODs and full-bandwidth recorders (Wildlife Acoustics, SM2M+) were deployed at two stations in an environment with fluctuating ambient noise for 42 days. Noise level and harbour porpoise (Phocoena phocoena) click trains simultaneously recorded on both loggers were compared. Overall, we found that porpoise click detections by the algorithm used to analyse full-band recorder data (Pamguard) paralleled detections by the C-POD. However, Pamguard detected significantly more clicks than the C-POD. A decrease in detections was seen for both loggers with increasing noise in the band 20 –160 kHz, in particular for levels above 100 dB re 1μPa rms. We also found that the Pamguard detection function changed the least over varying noise conditions when compared to the C-POD detectors. This study sheds light on the fact that inference of animal presence/absence or density that are based on echolocation cues (here, Porpoise Positive Minutes) shall account for the acoustic environments where probability of detecting signals may be affected by variability in ambient noise levels.     

Effects of signal duration on the recognition of masked communication signals by the grasshopper Chorthippus biguttulus

The detection and recognition of acoustic communication signals masked by noise was investigated in a grasshopper (Chorthippus biguttulus) whose auditory system exhibits only poor spectral resolution and therefore has to operate in the time domain. The signals of this species consist of numerous identical subunits that enable the receiver, in principle, to make repetitive measurements. We aimed at determining the maximum integration time in this species by using stimuli of different durations under increasing noise levels. As a criterion for recognition the typical phonotactic turning response of the males was evaluated, which is reliably triggered by a female song, and thus is a sensitive indicator for recognition of conspecific signals. When confronted with a long signal (1000 ms) males tolerated a 2.4 dB higher noise level as compared to a short signal (250 ms). Noise tolerance improved with increasing signal duration from 250 ms to 450 ms. Beyond this signal duration, however, no further improvement was observed, indicating an upper limit for temporal integration that corresponds to only five song subunits. The gain in noise tolerance had a slope of 2.7 dB per doubling duration, which corresponds to the expectation derived from an energy detector model (3 dB per doubling duration) rather than to the value expected from signal detection theory (1.5 dB per doubling duration).     

Animal signals: Ethological and games-theory approaches are not incompatible

Recent authors have contrasted the ‘traditional ethological approach’ to the study of animal signals with that derived from games theory. It is argued here that the ‘traditonal ethological approach’ they portray is not in keeping with main stream of ethological research on animal signals. In particular, it has not been generally assumed that the evolution of animal signals was based on selection for mutual benefit of actor and reactor, nor that signals carry precise information of what the actor will do next. A synthesis of the ethological and games-theory approaches is possible. It is suggested that many threat displays may signal ‘Will stay, but attack if provoked’ or ‘Will stay, but will flee if provoked’, and that the subsequent behaviour of the displaying bird depends in part on that of the reactor.     

Detection of Breast Cancer Based on Fuzzy Frequent Itemsets Mining

Background: Breast cancer, a type of malignant tumor, affects women more than men. About one third of women with breast cancer die of this disease. Hence, it is imperative to find a tool for the proper identification and early treatment of breast cancer. Unlike the conventional data mining algorithms, fuzzy logic based approaches help in the mining of association rules from quantitative transactions.Methods: In this study a novel fuzzy methodology IFFP (Improved Fuzzy Frequent Pattern Mining), based on a fuzzy association rule mining for biological knowledge extraction, is introduced to analyze the dataset in order to find the core factors that cause breast cancer. This method consists of two phases. During the first phase, fuzzy frequent itemsets are mined using the proposed algorithm IFFP. Fuzzy association rules are formed during the second phase, indicating whether a person belongs to benign or malignant. This algorithm is applied on WBCD (Wisconsin Breast Cancer Database) to detect the presence of breast cancer.Results: It is determined that the factor, Mitoses has low range of values on both malignant and benign and hence it does not contribute to the detection of breast cancer. On the other hand, the high range of Bare Nuclei shows more chances for the presence of breast cancer.Conclusion: Experimental evaluations on real datasets show that our proposed method outperforms recently proposed state-of-the-art algorithms in terms of runtime and memory usage.     

Choosing a Mate in a Cocktail Party‐Like Situation: The Effect of Call Complexity and Call Timing between Two Rival Males on Female Mating Preferences in the Túngara Frog Physalaemus pustulosus

Signal detection, recognition, and localization are hampered when multiple signalers coincide in time and space, a problem known as ‘cocktail party effect’. In many taxa, senders utter complex calls consisting of two or more elements which often vary in the ease with which they can be assessed in different signaling environments. Receivers’ selective attention to different cues may increase the probability of correctly assigning a signal to its source (localization) in face of conspecific interference. Túngara frogs, Physalaemus pustulosus, produce complex calls consisting of an initial whine, followed by zero up to seven broad‐banded, amplitude‐modulated chucks. Under ideal conditions (without interference or noise), females prefer whines followed by chucks over whines alone, but the preference is not linear; females do not discriminate between whines with one or two chucks. When whines lack chucks, call overlap elicits random responses in females, with no preference for leading calls. In this study, I explored the combined effect of call timing and call complexity on female preferences in a two‐choice paradigm—a simplification of the cocktail party scenario. I tested the hypothesis that the effect of call overlap can be reduced when the calls of one of the two rivals have chucks, specifically more chucks than those of the rival. I gave females a choice between whines alone and with chucks (one or two) presented at three time relations (alternated, abutted, and partially overlapped) and two emission orders (whine with less chucks leading and whine with more chucks leading). I found that the preference for one chuck over no chuck was preserved in all the experimental treatments, but when a w + 2chk preceded a w + chk, either overlapped or abutted, a preference existed for the whine with more chucks. Therefore, an interaction between call order and the number of chucks was obtained. The results only partially supported the hypothesis, and call order emerges as an opportunistic component of signaling in P. pustulosus.     

Fuzzy supervisory control of fed-batch baker's yeast process

Summary A fuzzy supervisory system for bioprocess control was developed, and applied to baker's yeast fermentation. The system was based on hierarchical bioprocess control with fuzzy phase recognition and separate fuzzy control of each process phase. A two-level knowledge base included rules both for the phase recognition and control. The system was tested by using experimental data of fed-batch baker's yeast cultivations and by process simulations.     

Analytical fuzzy approach to biological data analysis

The assessment of the physiological state of an individual requires an objective evaluation of biological data while taking into account both measurement noise and uncertainties arising from individual factors. We suggest to represent multi-dimensional medical data by means of an optimal fuzzy membership function. A carefully designed data model is introduced in a completely deterministic framework where uncertain variables are characterized by fuzzy membership functions. The study derives the analytical expressions of fuzzy membership functions on variables of the multivariate data model by maximizing the over-uncertainties-averaged-log-membership values of data samples around an initial guess. The analytical solution lends itself to a practical modeling algorithm facilitating the data classification. The experiments performed on the heartbeat interval data of 20 subjects verified that the proposed method is competing alternative to typically used pattern recognition and machine learning algorithms.     

Influence of amplitude modulated noise on the recognition of communication signals in the grasshopper Chorthippus biguttulus

The detection of acoustic communication signals in the presence of sinusoidally amplitude modulated noise was investigated in males of the grasshopper Chorthippus biguttulus. The auditory system of grasshoppers exhibits only poor spectral resolution. Hence, these animals are ideally suited to investigate noise tolerance in a system operating in the temporal domain. As a sensitive indicator for signal recognition the conspicuous phonotactic turning responses of males were recorded. The main result was that noise modulated at low frequencies (1.5-5 Hz) did not impair recognition compared to a unmodulated noise. With long stimuli even a moderate improvement of noise tolerance was observed, an effect that can probably be attributed to the existence of long troughs at low modulation frequencies during which the masking of the signal was reduced. Higher modulation frequencies (15-150 Hz), however, rendered detection and recognition increasingly difficult, due to a strong interference of the sound pulses of the masking noise with the syllable-pause structure of the species-specific signals. There are no indications for the operation of mechanisms analogous to comodulation masking release as found in vertebrates, nor for a spatial release from masking.     

Automatic decomposition of multichannel intramuscular EMG signals

We describe an automatic algorithm for decomposing multichannel EMG signals into their component motor unit action potential (MUAP) trains, including signals from widely separated recording sites in which MUAPs exhibit appreciable interchannel offset and jitter. The algorithm has two phases. In the clustering phase, the distinct, recurring MUAPs in each channel are identified, the ones that correspond to the same motor units are determined by their temporal relationships, and multichannel templates are computed. In the identification stage, the MUAP discharges in the signal are identified using matched filtering and superimposition resolution techniques. The algorithm looks for the MUAPs with the largest single channel components first, using matches in one channel to guide the search in other channels, and using information from the other channels to confirm or refute each identification. For validation, the algorithm was used to decompose 10 real 6-to-8-channel EMG signals containing activity from up to 25 motor units. Comparison with expert manual decomposition showed that the algorithm identified more than 75% of the total 176 MUAP trains with an accuracy greater than 95%. The algorithm is fast, robust, and shows promise to be accurate enough to be a useful tool for decomposing multichannel signals. It is freely available at http://emglab.stanford.edu.     

Auditory Perception of Voice Signals with Abnormal Sounding in Children

Invariant and noise-proof speech understanding is an important human ability, ensured by several mechanisms of the audioverbal system, which develops parallel to mastering linguistic rules. It is a fundamental problem of speech studies to clarify the mechanisms of this understanding, especially their role in the speech development. The article deals with of the regularities of auditory word recognition in noise by preschool children (healthy and with speech development disorders) and patients with cochlear implants. The authors studied the recognition of words using pictures (by children) and verbal monitoring, when the subjects were stimulated by isolated words with one or all syllables in noise. The study showed that children's ability to perceive distorted words develops in ontogeny and is closely related to the development of mental processes and mastering linguistic rules. The data on patients with cochlear implants also confirmed the key role of the central factors in understanding distorted speech.