Machine Learning Guides Peptide Nucleic Acid Flow Synthesis and Sequence Design

Abstract

Peptide nucleic acids (PNAs) are potential antisense therapies for genetic, acquired, and viral diseases. Efficiently selecting candidate PNA sequences for synthesis and evaluation from a genome containing hundreds to thousands of options can be challenging. To facilitate this process, this work leverages machine learning (ML) algorithms and automated synthesis technology to predict PNA synthesis efficiency and guide rational PNA sequence design. The training data is collected from individual fluorenylmethyloxycarbonyl (Fmoc) deprotection reactions performed on a fully automated PNA synthesizer. The optimized ML model allows for 93% prediction accuracy and 0.97 Pearson's r. The predicted synthesis scores are validated to be correlated with the experimental high-performance liquid chromatography (HPLC) crude purities (correlation coefficient R 2 = 0.95). Furthermore, a general applicability of ML is demonstrated through designing synthetically accessible antisense PNA sequences from 102 315 predicted candidates targeting exon 44 of the human dystrophin gene, SARS-CoV-2, HIV, as well as selected genes associated with cardiovascular diseases, type II diabetes, and various cancers. Collectively, ML provides an accurate prediction of PNA synthesis quality and serves as a useful computational tool for informing PNA sequence design.

Computer vision as a new paradigm for monitoring of solution and solid phase peptide synthesis

“…23 For monitoring of liquid phase peptide synthesis, Livingston's team have reported a powerful advance in UHPLC-MS. 24 Gómez-Bombarelli and Pentelute have demonstrated the value of UV-vis analysis of in-ow Fmoc deprotection reactions to generate data-rich input to build predictive deep learning insights for Fmoc deprotection efficiency. 15,25 Otake and co-workers have demonstrated the value of in-line near infrared (NIR) ow cells as a means of tracking liquid phase peptide synthesis in ow. 26 We used our recently-developed imaging kinetics soware, Kineticolor, 27,28 to provide camera-enabled ex situ (non-contact) reaction monitoring (Fig.…”

Machine learning for revealing the relationship between the process–structure–properties of polypropylene in-reactor alloys

“…14 Consequently, machine learning has been widely used in structure determination, performance prediction and new material discovery. 15,16 Bhowmik et al employed Decision Tree and principal component analysis methods to successfully predict the specific heat capacity of polymers using various structural descriptors, including bonds, angles, atoms and molecular weights. 17 Joo et al developed prediction models for the physical properties of polypropylene composites, employing three machine learning methods: multiple linear regression, deep neural network and Random Forest.…”

Discovery of thermosetting polymers with low hygroscopicity, low thermal expansivity, and high modulus by machine learning

“…16 Machine learning (ML) tools can be applied to solve many problems in chemical science and materials science, including mapping structure and property relationships, [19][20][21][22] discovering new compounds and materials, [23][24][25][26][27][28][29][30] recommending feasible chemical synthesis pathways, [31][32][33] and guiding automated experiments. [34][35][36] ML tools are data-driven; therefore, highquality, diverse, and well-organized data are the foundation of these studies. 8,37 However, the sparse, small, and difficult-toextract nature of polymer data severely restricts the application of these tools and even the development of polymer informatics.…”