Generating experimentally unrelated target molecule-binding highly functionalized nucleic-acid polymers using machine learning

Abstract

In vitro selection queries large combinatorial libraries for sequence-defined polymers with target binding and reaction catalysis activity. While the total sequence space of these libraries can extend beyond 1022 sequences, practical considerations limit starting sequences to ≤~1015 distinct molecules. Selection-induced sequence convergence and limited sequencing depth further constrain experimentally observable sequence space. To address these limitations, we integrate experimental and machine learning approaches to explore regions of sequence space unrelated to experimentally derived variants. We perform in vitro selections to discover highly side-chain-functionalized nucleic acid polymers (HFNAPs) with potent affinities for a target small molecule (daunomycin KD = 5–65 nM). We then use the selection data to train a conditional variational autoencoder (CVAE) machine learning model to generate diverse and unique HFNAP sequences with high daunomycin affinities (KD = 9–26 nM), even though they are unrelated in sequence to experimental polymers. Coupling in vitro selection with a machine learning model thus enables direct generation of active variants, demonstrating a new approach to the discovery of functional biopolymers.

A Sensitive and Nonoptical CRISPR Detection Mechanism by Sizing Double‐Stranded λ DNA Reporter

“…Although the interaction of self‐assembled DNA/RNA nanostructures to other small molecules, cells, and proteins (including Cas protein [27] ) has been studied previously,[ 27 , 28 ] the origin of binding affinity between Cas12a and λ DNA requires further investigation. In this perspective, a modeling approach [29] could help identify the specific regions of λ DNA that show an affinity to Cas12a in future.…”

An explainable machine-learning approach for revealing the complex synthesis path–property relationships of nanomaterials

“…Recently, with the development of artificial intelligence, machine learning (ML) has shown excellent application potential in natural sciences. For instance, ML-driven approaches have been applied to molecular recognition, 13–16 materials design and discovery, 17–20 and synthesis reaction prediction. 21–24 However, the low interpretability of ML models obscures the distribution and importance of parameters and the interaction between parameters, which affects the credibility and dependency of ML models.…”

BinderSpace: A Package for Sequence Space Analyses for Datasets of Affinity-Selected Oligonucleotides and Peptide-Based Molecules

“…The types of molecules comprising the libraries can include the oligonucleotides themselves, functionalized nucleic acid polymers 4 , small molecules 5,6 , linear, cyclic or chemically modified peptides [7][8][9][10][11] , and novel hybrid molecules 12 . A method for selecting target binders from libraries of oligonucleotides of varying sequences 13,14 is called SELEX (systematic evolution of ligands by exponential enrichment).…”

“…With the advent of artificial intelligence (AI) approaches, there is a large interest in using experimental selection datasets for training machine learning (ML) models to predict oligonucleotides, peptide-based and other molecules with high affinity for targets 4,22 or the highest ability to impart functional response to the target, such as the induction of a fluorescence signal [23][24][25] . Bioinformatics analyses can be used to understand sequence composition of experimental datasets and also to assess the sequence patterns in sets of molecules predicted to have the highest affinity for targets by the ML models.…”