Predicting Analyte Concentrations from Electrochemical Aptasensor Signals Using LSTM Recurrent Networks

Abstract

Nanomaterial-based aptasensors are useful devices capable of detecting small biological species. Determining suitable signal processing methods can improve the identification and quantification of target analytes detected by the biosensor and consequently improve the biosensor’s performance. In this work, we propose a data augmentation method to overcome the insufficient amount of available original data and long short-term memory (LSTM) to automatically predict the analyte concentration from part of a signal registered by three electrochemical aptasensors, with differences in bioreceptors, analytes, and the signals’ lengths for specific concentrations. To find the optimal network, we altered the following variables: the LSTM layer structure (unidirectional LSTM (LSTM) and bidirectional LSTM (BLSTM)), optimizers (Adam, RMSPROP, SGDM), number of hidden units, and amount of augmented data. Then, the evaluation of the networks revealed that the highest original data accuracy increased from 50% to 92% by exploiting the data augmentation method. In addition, the SGDM optimizer showed a lower performance prediction than that of the ADAM and RMSPROP algorithms, and the number of hidden units was ineffective in improving the networks’ performances. Moreover, the BLSTM nets showed more accurate predictions than those of the ULSTM nets on lengthier signals. These results demonstrate that this method can automatically detect the analyte concentration from the sensor signals.

Anomaly Detection for Sensor Signals Utilizing Deep Learning Autoencoder-Based Neural Networks

“…Since the main focus of this study is to find abnormal signals by utilizing the autoencoder networks, we avoided explaining all characteristics and features of these biosensors. A detailed description of these biosensors’ components and their sensing protocols can be found in our previous article [ 20 ]. In addition, comprehensive information on the 35-mer adenosine biosensor, such as fabricating the transistor and functionalizing the aptamer, can be found in [ 21 ].…”

“…In our previous work [ 20 ], the anomalies, called contextual outliers, were detected by data visualization and labelled by the data collectors as normal, marginal, no-sensing signals, and broken transistors. Then, only those signals labelled as normal were kept for making a deep learning prediction model.…”

“…They assigned the following labels on the signals or segments: normal, marginal, no sense, and broken transistor. In our previous work [ 20 ] and in this paper, those signals with marginal, no sense, and broken transistor labels were considered anomalies.…”

“…The input gate controls the values that need to be updated, and the cell candidate adds information to the LSTM memory. The forget gate selects the contributory fraction of information to the output [ 20 , 29 ].…”

“…This paper was motivated by our previous work [ 20 ], where we developed an LSTM-based prediction model for classifying the biosensors’ analyte concentration. The outlier detection was addressed by data visualization and labelling with expert knowledge.…”