Molecular convolutional neural networks with DNA regulatory circuits

Abstract

Complex biomolecular circuits enable cells with intelligent behavior for survival before neural brains evolved. Synthesized DNA circuits in liquid phase developed as computational hardware can perform neural-network-like computation that harness the collective properties of complex biochemical systems, however the scaling up in complexity remains challenging to support more powerful computation. we present a systematic molecular implementation of the convolutional neural network (ConvNet) algorithm with synthetic DNA regulatory circuits based on a simple DNA switching gate architecture. We experimentally demonstrated that a DNA-based ConvNet based on shared-weight architecture of a 3×6 sized kernel can simultaneously implement parallel multiply-accumulate (MAC) operations for 144 bits inputs and recognize patterns up to 8 categories autonomously. Furthermore, it can connect with another DNA circuits to construct hierarchical networks, which can recognize patterns up to 32 categories with a two-step classi cation approach of performing coarse classi cation on language (Arabic numerals, Chinese oracles, English alphabets and Greek alphabets) and then classifying them into speci c handwritten symbols. With a simple cyclic freeze/thaw approach, we can decrease computation time from hours to minutes. Our approach shows great promise in the realization of high computing power molecular computer with ability to classify complex and noisy information.

Molecular Computation for Molecular Classification

“…[ 121 ] B) Molecular convolutional network for the recognition of symbols. [ 123 ] C) Classifier of bacterial versus viral infection from host RNA transcripts. [ 124 ] D) Neural sensor of lung cancer patient from 4 input circulating miRNAs.…”

DNA strand displacement based computational systems and their applications

“…The essence of the convolution operation is matrix multiplication. Xiong et al (2022) take the same matrix multiplication strategy as Genot et al (2013). They designed the Metrix operation DSD system, including the multiplication of two matrices, matrix Addition, and matrix Subtraction.…”

“…What's more, integrated DSD circuits can function as artificial neural networks (ANN) and perform machine learning (ML) algorithms. The Multilayer Perceptron (MLP) (Arredondo and Lakin, 2022), Support Vector Machine (SVM) (Lopez et al, 2018), Hopfield network , and Convolutional Neural Network (CNN) (Xiong et al, 2022) had been architected via DSD reactions.…”

“…The right part shows the convolution results. Reproduced with permission(Xiong et al, 2022). Copyright 2022, Nature Publishing Group.…”

Intramolecular and Intermolecular Hole Delocalization Rules the Reducer Character of Isolated Nucleobases and Homogeneous Single-Stranded DNA

“…For this reason, intense research has been carried out around its structure and functions. Moreover, DNA strands can also be employed for many technological purposes, 1 including their use as nanowires, 2,3 DNA-templated synthesis for new materials, 4 DNA computation, 5,6 and molecular detection [7][8][9][10][11][12][13] , among others. Regarding the latter, DNA-based biosensors are becoming nowadays promising devices in sensor research, where the detection process is usually carried out by optical or electrochemical techniques.…”