Projection of High-Dimensional Genome-Wide Expression on SOM Transcriptome Landscapes

The self-organizing maps portraying has been proven to be a powerful approach for analysis of transcriptomic, genomic, epigenetic, single-cell, and pathway-level data as well as for “multi-omic” integrative analyses. However, the SOM method has a major disadvantage: it requires the retraining of the entire dataset once a new sample is added, which can be resource- and time-demanding. It also shifts the gene landscape, thus complicating the interpretation and comparison of results. To overcome this issue, we have developed two approaches of transfer learning that allow for extending SOM space with new samples, meanwhile preserving its intrinsic structure. The extension SOM (exSOM) approach is based on adding secondary data to the existing SOM space by “meta-gene adaptation”, while supervised SOM portrayal (supSOM) adds support vector machine regression model on top of the original SOM algorithm to “predict” the portrait of a new sample. Both methods have been shown to accurately combine existing and new data. With simulated data, exSOM outperforms supSOM for accuracy, while supSOM significantly reduces the computing time and outperforms exSOM for this parameter. Analysis of real datasets demonstrated the validity of the projection methods with independent datasets mapped on existing SOM space. Moreover, both methods well handle the projection of samples with new characteristics that were not present in training datasets.

Using dynamic time warping self-organizing maps to characterize diurnal patterns in environmental exposures

Advances in measurement technology are producing increasingly time-resolved environmental exposure data. We aim to gain new insights into exposures and their potential health impacts by moving beyond simple summary statistics (e.g., means, maxima) to characterize more detailed features of high-frequency time series data. This study proposes a novel variant of the Self-Organizing Map (SOM) algorithm called Dynamic Time Warping Self-Organizing Map (DTW-SOM) for unsupervised pattern discovery in time series. This algorithm uses DTW, a similarity measure that optimally aligns interior patterns of sequential data, both as the similarity measure and training guide of the neural network. We applied DTW-SOM to a panel study monitoring indoor and outdoor residential temperature and particulate matter air pollution (PM2.5) for 10 patients with asthma from 7 households near Salt Lake City, UT; the patients were followed for up to 373 days each. Compared to previous SOM algorithms using timestamp alignment on time series data, the DTW-SOM algorithm produced fewer quantization errors and more detailed diurnal patterns. DTW-SOM identified the expected typical diurnal patterns in outdoor temperature which varied by season, as well diurnal patterns in PM2.5 which may be related to daily asthma outcomes. In summary, DTW-SOM is an innovative feature engineering method that can be applied to highly time-resolved environmental exposures assessed by sensors to identify typical diurnal (or hourly or monthly) patterns and provide new insights into the health effects of environmental exposures.

Algorithmic Study of the Characteristics of Electrocardiograph Signals in Patients with Coronary Heart Disease

This work aimed to analyze the electrocardiogram (ECG) characteristics and signal classification of patients with coronary heart disease (CHD) diagnosed by coronary angiography, so as to provide a theoretical basis for the clinical adoption of ECG images. 106 patients with CHD who were admitted to the XXX hospital from January 15, 2019, to May 30, 2020, underwent coronary intervention therapy, and their ECG indicators were recorded during the operation. Then, the LetNet-SoM algorithm designed in this work, as well as the traditional algorithms GoogLeNet and SqueezeNet, was applied to the patient’s ECG classification. It was found that part of ECG wave (QRS) and corrected Q-T interval (QTC) of patients after treatment were higher than those before treatment ( P < 0.05 ), but PR interval, RR interval, Tpeak-Tend (TpTe) interval, and QT interval were not substantially different from those before treatment ( P > 0.05 ). The diagnostic accuracy, sensitivity, and specificity of LetNet-SoM algorithm for patients with CHD were better than those of traditional algorithms, with evident difference ( P < 0.05 ). The classification time of LetNet-SoM algorithm was lower in contrast to that of traditional algorithms, and the difference was also notable ( P < 0.05 ). The R wave and T wave indicators of patients after treatment were higher than before treatment, with notable difference ( P < 0.05 ). The difference between the patient’s S wave indicator before and after treatment was not statistically significant ( P > 0.05 ). The positive rate of S wave amplitude, QRS, and QTC was 68.15%, 60.52%, and 51.36%, respectively. In short, the LetNet-SoM algorithm designed based on lightweight neural network had excellent performance in classification and diagnosis of ECG, and it had the value of further popularization and application. The ECG signals were important indicators in the diagnosis of CHD, among which the S wave amplitude, QRS, and QTC were the most sensitive ones.

Implementasi Data Mining untuk mengelompokkan Pasien Rawat Jalan dan Rawat Inap Asuransi Kesehatan

Patient Visits Outpatient and inpatient insurance at Class C Hospitals is increasing from year to year. Increased visits to insurance patients will have an impact on the inpatient and outpatient health services provided. From the increase in patient visits, the data owned by the hospital is increasingly abundant. The data can be used to explore knowledge, find certain patterns. To explore knowledge about Inpatient and Outpatient Insurance patients, data mining clustering techniques are used with the Self Organizing Map (SOM) algorithm using R Studio tools. Clustering technique with the implementation of the Self Organizing Map (SOM) algorithm is a technique for grouping data based on certain characteristics which are then mapped into areas that resemble map shapes. The CRISP-DM method is used in this study to perform the stages of the data mining process. The results obtained from the implementation of clustering with the Self Organizing Map (SOM) algorithm are obtained 2 clusters representing dense areas and non-congested areas. Dense areas are represented by Internal Medicine Clinic, Surgery Clinic, Eye Clinic, Hemodialysis, Melati Room, Orchid Room, Bougenville Room, Flamboyan Room. Non-crowded areas are represented by General Clinics, Dental Clinics, Obstetrics and Gynecology Clinics, Children's Clinics, Mawar Room and Soka Room

S-SOM v1.0: a structural self-organizing map algorithm for weather typing

This study proposes a novel structural self-organizing map (S-SOM) algorithm for synoptic weather typing. A novel feature of the S-SOM compared with traditional SOMs is its ability to deal with input data with spatial or temporal structures. In detail, the search scheme for the best matching unit (BMU) in a S-SOM is built based on a structural similarity (S-SIM) index rather than by using the traditional Euclidean distance (ED). S-SIM enables the BMU search to consider the correlation in space between weather states, such as the locations of highs or lows, that is impossible when using ED. The S-SOM performance is evaluated by multiple demo simulations of clustering weather patterns over Japan using the ERA-Interim sea-level pressure data. The results show the S-SOM's superiority compared with a standard SOM with ED (or ED-SOM) in two respects: clustering quality based on silhouette analysis and topological preservation based on topological error. Better performance of S-SOM versus ED is consistent with results from different tests and node-size configurations. S-SOM performs better than a SOM using the Pearson correlation coefficient (or COR-SOM), though the difference is not as clear as it is compared to ED-SOM.

Combination of Computational Techniques to Obtain High-Quality Gelatin-Base Gels from Chicken Feet

With the increasing global population, it has become necessary to explore new alternative food sources to meet the increasing demand. However, these alternatives sources should not only be nutritive and suitable for large scale production at low cost, but also present good sensory characteristics. Therefore, this situation has influenced some industries to develop new food sources with competitive advantages, which require continuous innovation by generating and utilising new technologies and tools to create opportunities for new products, services, and industrial processes. Thus, this study aimed to optimise the production of gelatin-base gels from chicken feet by response surface methodology (RSM) and facilitate its sensorial classification by Kohonen’s self-organising maps (SOM). Herein, a 22 experimental design was developed by varying sugar and powdered collagen contents to obtain grape flavoured gelatin from chicken feet. The colour, flavour, aroma, and texture attributes of gelatines were evaluated by consumers according to a hedonic scale of 1–9 points. Least squares method was used to develop models relating the gelatin attributes with the sugar content and collagen mass, and their sensorial qualities were analysed and classified using the SOM algorithm. Results showed that all gelatin samples had an average above six hedonic points, implying that they had good consumer acceptance and can be marketed. Furthermore, gelatin D, with 3.65–3.80% (w/w) powdered collagen and 26.5–28.6% (w/w) sugar, was determined as the best. Thus, the SOM algorithm proved to be a useful computational tool for comparing sensory samples and identifying the best gelatin product.

Combining expert knowledge and machine-learning to classify herd types in livestock systems

A detailed understanding of herd types is needed for animal disease control and surveillance activities, to inform epidemiological study design and interpretation, and to guide effective policy decision-making. In this paper, we present a new approach to classify herd types in livestock systems by combining expert knowledge and a machine-learning algorithm called self-organising-maps (SOMs). This approach is applied to the cattle sector in Ireland, where a detailed understanding of herd types can assist with on-going discussions on control and surveillance for endemic cattle diseases. To our knowledge, this is the first time that the SOM algorithm has been used to differentiate livestock systems. In compliance with European Union (EU) requirements, relevant data in the Irish livestock register includes the birth, movements and disposal of each individual bovine, and also the sex and breed of each bovine and its dam. In total, 17 herd types were identified in Ireland using 9 variables. We provide a data-driven classification tree using decisions derived from the Irish livestock registration data. Because of the visual capabilities of the SOM algorithm, the interpretation of results is relatively straightforward and we believe our approach, with adaptation, can be used to classify herd type in any other livestock system.