A scientist working in the field of DNA nanotechnology can be very successful without knowing how to draw the chemical structure of DNA. Indeed, from a programming perspective, DNA is stunningly simple: a string of bits coding two types of interactions, i.e., a digital polymer (Fig. 1, left). But from a chemistry perspective, DNA is exceedingly complex: heterocycles, sugars, and phosphates are organized via a diverse arsenal of chemical bonds—covalent, hydrogen, ionic, and π–π stacking—to give just the right structure enabling robust recognition by many complex biomolecules for reading, copying, evolution, and a multitude of other biological functions (Fig. 1, right).

Fig. 1
2 illustrations. Illustration 1 has 2 parallel lines with 0 and 0 prime, and 1 and prime are arranged vertically and connected with dashed lines. Illustration 2 is the structure of D N A with fused aromatic rings, hydrogen bonding and different functional groups.

DNA from the perspective of a molecular programmer (left) and a chemist (right)

Full size image

This specific form of DNA given to us by evolution imposes significant constraints on what is possible with DNA nanotechnology. For example, G-quadruplex formation limits sequences available for strand displacement, and cleavage by nucleases limits in vivo utility of DNA-only structures. Nevertheless, DNA remains the most powerful molecule for molecular programming due to the simplicity of its programming rules and its practical advantages including: (i) fast and cheap custom synthesis (e.g., by integrated DNA technologies and Twist Bio), (ii) mature design and modeling tools (e.g., scadnano [1], Peppercorn [2], MagicDNA [3], NUPACK [4], and OxDNA [5]), (iii) ability to amplify copy number (e.g., by PCR), and (iv) fast and cheap sequencing (e.g., by Pacific Bio and Oxford Nanopore). In this paper, I will describe three new digital polymers that can enable functions not achievable with DNA while retaining the essential information-bearing properties and some practical advantages of DNA.

An astute reader may be wondering, “this is a very ambitious proposal. People have worked on DNA alternatives for many years. What about all the existing polymers such as PNA, GNA, and other XNAs?” Yes, many people have worked on DNA alternatives for decades exploring questions ranging from pure curiosity about molecular origins of life [6] to very practical goals of improving microarray sensors [7]. Typically, these alternatives have been focused on creating molecules possessing a specific attribute of DNA, most notably (a) transcription and translation by enzymes [8], (b) ability to recognize DNA sequences while being more stable [9], and (c) information storage [10]. These alternatives include GNA (a DNA analog in which sugar is replaced with a more stable glycerol [11]), phosphorothioate DNA (a DNA analog in which oxygen in phosphate is replaced with sulfur [12]), PNA (a DNA analog in which sugar is replaced with a neutral peptide-like backbone [13]), and L-DNA (a DNA analog with opposite chirality to DNA [14]). It may seem surprising, but no one, to my knowledge, has tried to redesign DNA to enlarge its functionality for the purposes of DNA computing or DNA nanotechnology. As the new digital polymer designs below will demonstrate, without the constraint to bind natural DNA or be processed by natural enzymes, we have much more freedom in molecular design.

Before describing the new polymer designs, it is instructive to first consider what desirable functions of a digital polymer DNA currently lacks, and how these might be achievable by existing approaches without building an entirely new polymer. Significant limitations of DNA, summarized in Table 1, include low stability, limited chemical functionality with 4-word alphabet, and lack of bio-orthogonality. Generally, to overcome these limitations, chemical modifications or non-DNA coating are used. However, these modifications are frequently hard to implement for non-chemists and often disrupt the desired structure of DNA.

Table 1 Desired properties/abilities and approach to achieve them with DNA-based structures
Full size table

While combining all the above functions in one polymer is neither feasible nor necessary, I propose three new polymers below that draw from this pool of functions to create DNA analogs that better serve specific tasks. For each polymer, I will specifically discuss redesign of the recognition elements and backbone, the new functions this enables, and synthesis. The first polymer, NP1, marginally redesigns DNA bases while introducing a simpler and more modular α-peptide backbone. This should expand its chemical vocabulary and programmability, among other advantages. A second, more ambitious polymer design, NP2, proposes a recognition code based entirely on natural peptides. This should also increase chemical functionality while enabling scalable production in vivo. A third polymer, NP3, aims to make a covalent recognition code for applications requiring very stable architectures. This should yield polymers with ultimate stability, e.g., for making portable devices. I hope that the ideas explored in these three designs can lead to an extended toolkit for molecular programming that will enable nanotechnologies currently impractical with DNA.

1 New Polymer 1 (NP1)

The design of NP1 builds on the successes of DNA, proteins, and peptide nucleic acid (PNA) while adding new functions unattainable by any single class of these molecules such as higher stability and larger design space. First, I propose to “fix” DNA bases for a slimmer and stronger code. Then, I explain why an α-peptide backbone is better than any other backbone for a digital polymer, especially when it comes to scalable production of the new polymer. Finally, I provide an overview of related existing approaches and discuss ways to overcome potential pitfalls.

Recognition elements. The recognition between two strands of NP1 is based on unmodified DNA bases (T and C) and modified DNA bases (A and G). First, to enable denser molecular pegboards, I propose to make the double helix slimmer by replacing purine bases with pyrimidine, effectively chopping off the 5-membered ring (Fig. 2, blue fragments; compare to DNA). One concern is that this “slimming surgery” may disrupt the balance between hydrophobic and hydrophilic interactions that has been finely tuned during molecular evolution, thus destabilizing the helix. Another concern is that this design goes against the size complementarity hypothesis, which states that large purines must pair with small pyrimidines. In fact, however, “skinny” DNA where purines are replaced with pyrimidines preserves its structure and even gains stability [28]. Another advantage of removing the 5-membered heterocycle is that this eliminates guanine’s ability to form quadruplexes, a big nuisance for some DNA applications such as information retrieval in G-rich DNA sequences for data storage.

Fig. 2
4 molecular structures of D N A, P N A, N P 1. and N P 1 with peptoid A C G T. The structures have fused rings with hydrogen bonding and different functional groups.

Structures of DNA, PNA, new polymer 1 (NP1), and NP1 with peptoid ACGT analogs

Full size image

To improve the stability of A–T pairs, I propose to add an amino group to the modified adenine (Fig. 2, red). This converts an asymmetric double hydrogen bond to a symmetric triple hydrogen bond, which has been shown to increase duplex stability while precluding stray recognition by some biomolecules normally binding to the minor groove [8]. To preserve hydrogen-bonding motifs, I propose to move the nitrogens in the new pyrimidines away from the ring location linked to the backbone. The reason I retain the nitrogens is that their electron withdrawing effect is important to maintain the same energy of lowest unoccupied molecular orbital (LUMO) localized on the lone pair of the hydrogen-bonding amino group in position 2 of the new pyrimidine analog of adenine. One can also achieve a similar electron withdrawing effect by introducing a nitro group in position 6 of the ring [28]. Another reason to move nitrogen to the new position 3 is that it is easier to synthesize such a base compared to one with carbon.

Backbone. Unlike PNA, NP1 organizes hydrogen-bonding elements on a simpler yet more versatile natural α-amino acid backbone. This can potentially allow adaptation of not only existing chemical but also biological peptide syntheses to produce NP1. Adding a canonical amino acid after every recognition bearing non-canonical amino acid (Fig. 2, colorful Rs) also allows a great expansion of the chemical functionality available to NP1 compared to DNA or PNA. A larger chemical space afforded by more diverse and numerous amino acid residues compared to nucleotides is likely one of the main reasons why Nature has eventually chosen peptides and not polynucleotides to build most molecular machines.

To keep the initial designs as close to DNA as possible, I propose to place hydrogen-bonding bases on every second amino acid, which should give a base spacing very close to the ones in DNA and PNA (PNA forms a double helix with 18 bases per period versus 10.5 for DNA [29].) Unlike PNA, NP1 is designed to be chiral by adding chiral (L-) ACGT amino acids and natural L-amino acids. As a reminder, PNA lacks chirality due to the rapid inversion of the tertiary prochiral nitrogens. This inversion leads to racemic mixtures of left- and right-handed helices, as well as promiscuity in C⟶N polarity of the peptide backbone, an equivalent of 5′⟶3′ polarity of the sugar backbone that enforces the antiparallel requirement for DNA helices. While it is possible to add chirality to the PNA backbone by synthesizing γ-modified PNA, NP1 is more modular compared to such a γ-modified PNA, because NP1 requires only 24 monomers (four ACGT analogs and 20 α-amino acids) to enable more design freedom than possible with 80 γ-modified PNA monomers (4 × 20). Additionally, while PNA’s neutrality makes it challenging to deliver in vivo, NP1’s hydrophilicity may be tuned by incorporating a variety of hydrophilic and hydrophobic amino acid residues enabling applications in media with a range of lipophilicities from water to hexane. However, care should be taken to avoid combinations which are unsuitable, including large regions of highly polar side chains or unbroken polar-non-polar alternating regions to avoid amphiphilic or sheet-forming behavior.

Looking at DNA through the eyes of a chemist also presents a tantalizing opportunity to explore the design space of two-stranded helical molecules. For example, in NP1, nucleobase analogs can be integrated more sparsely (with more natural amino acids inserted between the ACGT analogs compared to just a single amino acid in Fig. 2), allowing natural biologically-active peptide fragments to be organized with a precision and complexity infeasible using current protein engineering approaches [30]. Also, by varying the number and the type (α, β, γ, δ) of amino acids, it should be possible to control the helical pitch, i.e., the angle between two adjacent stacks, which is 34.3° for DNA and 20° for PNA. In an extreme case, it may be possible to reduce the rotation per base pair to 0° by directly linking two nucleobase analogs (skipping a canonical amino acid). This is because the distance between two adjacent nucleobase analogs attached to a fully stretched α-amino acid peptide backbone is 3.5 Å, which means that the scaffold almost does not need to twist for the bases to reach the typical stacking distance of 3.4 Å.

Synthesis. To realize its practical potential, a new digital polymer should be amenable to automated synthesis, allowing efficient production of custom sequences as can be done now for DNA, peptides, and oligosaccharides [31]. Initial efforts may most profitably be dedicated to the synthesis of the four amino acids nucleobase analogs that are compatible with the standard automated peptide or peptoid syntheses [32]. Next, the effect of sequence and other design elements on the structure of the double-stranded helix and simple multi-strand constructs using nuclear magnetic resonance (NMR), X-ray crystallography, and atomic force microscopy (AFM) should be investigated.

Once optimal structures are elucidated, efforts should be focused on developing larger-scale sustainable synthesis. One approach can be to adapt natural cellular protein synthesis machinery to produce large amounts of NP1; this will leverage the last two decades of rapid progress in techniques for co-translational incorporation of unnatural amino acids into proteins produced in cells via genetic code expansion [33]. Finally, directed evolution may be used to modify components of the translation machinery such as synthetase/tRNA pairs to selectively incorporate the four new amino acids with A, C, G, T analog residues first in cell-free media and then in cells [34].

Existing approaches to design a simple non-covalent recognition code. Much work has been done to create digital polymers with recognition between strands based on hydrogen bonds and other non-covalent interactions [35,36,37,38]. However, a DNA-like polymer with robust recognition of DNA has not yet been reported, partially because creating such a digital polymer for the purposes of DNA nanotechnology was not the main reason for these efforts.

Potential pitfalls and ways to overcome them. Experts on protein structure may have a valid concern that instead of the desired double helix, NP1 may form α-helix, β-sheet, or other motifs common in proteins due to the presence of carbonyls and NH groups. I expect that the three locally organized hydrogen bonds in the base pairs will be more favorable than the two relatively remote hydrogen bonds of α-helices or β-sheets. Also, proline, an amino acid that lacks NH, can be sparsely added to break up α-helices or β-sheets. Yet another way to minimize spurious hydrogen bonding is to use peptoid versions of ACGT where the base is attached to the nitrogen (Fig. 2). As a note, PNA has the capacity to form spurious NH···OC hydrogen bonds (twice as few as α-peptides but the same number as the peptoid version of NP1) but still prefers to form a double helix. However, I propose to keep regular α-amino acid monomers for non-ACGT monomers instead of fully switching to peptoids to retain the desired chirality and higher chance for in vivo synthesis. In planning in vivo applications, one should keep in mind that while gaining stability to nucleases NP1 will become susceptible to proteases. In addition, one should consider that NP1 may bind DNA especially in a single stranded form due to the similarity of its recognition elements to DNA bases.

2 New Polymer 2 (NP2)

Design. NP1 requires developing synthesis and incorporation of synthetic DNA base analogs. Is it possible to avoid this and design a new DNA-like polymer entirely from natural amino acids? The design of NP2 also relies on the natural α-amino acid peptide backbone. However, instead of familiar ACGT analogs, this design seeks to construct recognition elements from natural amino acid side chains (Fig. 3). Hydrophobic interactions such as V–L or F–W or single hydrogen bonds in pairs such as S–T or C–Y are too weak and not specific to provide a reliable code. One excellent candidate is R–D pair that is maintained by two hydrogen bonds and electrostatic attraction. This interaction is likely even stronger than the G–C hydrogen bond. In principle, it is possible to construct a digital polymer using just a single heterophilic interaction (i.e., binding partners are different) such as G–C or R–D, but this would substantially increase possible spurious interactions and decrease information density for the same length compared to a polymer with two heterophilic interactions such as DNA. Therefore, it is desirable to introduce another specific interaction that is as orthogonal to R-D as possible. Since the only other available heterophilic interaction—K–E—also involves a carboxylic acid (E, on a slightly longer leash than D) and will likely bind to R, I suggest using a homophilic interaction (i.e., binding partners are the same) such as N–N or Q–Q or Q–N. This interaction is also maintained by two hydrogen bonds, but the partners are not charged as in the R-D pair. The lack of charges should bring HOMO and LUMO of binding partners closer, making this interaction more orthogonal to R–D. Among the three possible interactions (N–N or Q–Q or Q–N), I show N–N in Fig. 3, because according to preliminary molecular dynamics calculations performed in my group, N–N gives more stable double-stranded complexes compared to Q–Q, Q–N, and N–Q. R, D, and G essentially comprise a three-letter code which will likely have more spurious interactions than a four-letter code (A, T, G, C). My group is currently studying a series of complexes such as shown in Fig. 3 (R = G and L) by NMR, AFM, and X-ray diffraction. P is inserted to disrupt potential α-helices or β-sheets as discussed for NP1.

Fig. 3
20 structures of Glycine, Alanine, Valine, Leucine, Isoleucine, Methionine, Phenil alanine, Tryptophan, Proline, Serine, Threonine, Cysteine, Tyrosine, Aspargine, Glutamine, Aspartate, Glutamate, Lysine,Argininine, and Histidine on the left. The structure of N P 2 with hydrogen bonds on the right.

Structures of 20 essential amino acids (left) and an example structure of New Polymer 2 duplex (right) made with a self-complementary strand. Red Rs can potentially be various side chains from the amino acids on the left

Full size image

Existing approaches to design a simple recognition code with peptides. The idea that it is possible to construct a recognition code similar in simplicity to A–T and G–C of DNA with natural amino acids may seem preposterous to some. “If there had been such a simple peptide motif, Nature (not the journal) would surely have found it by now. If not Nature, then David Baker”—an active reader may think. And the reader would be partially right: Some recognition elements have indeed been identified both by stochastic natural evolution and more rational protein design. However, these motifs are still not as simple, programmable, and practical as A–T and G–C of DNA, as I explain next.

Some obvious hydrogen-bonding recognition analogs of A-T and G-C mentioned above are R–D, R–E, K–D, K–E, N–N, Q–Q, and Q–N with their two hydrogen bonds. Protein database (PDB) search indeed reveals quite frequent intra- and interpeptide contacts with shorter than Van-der-Waals radii distances for these pairs indicating bonding. So, Nature does use them, but it does so very rarely compared to NH-CO hydrogen bonds and hydrophobic residue interactions that dominate protein folding landscapes. Learning from the wealth of protein structures in PDB, protein designers have made huge progress in predicting unknown structures (e.g., with Rosetta [39] and AlphaFold [40]) and designing new ones [41]. However, even the simplest recognition codes elucidated so far are still not as compact and modular as A–T and G–C of DNA, limiting their practical use [42].

The great wealth of natural PDB structures combined with machine learning algorithms is very powerful, but unlikely to uncover a conceptually new code. The current protein world is biased to represent structures achievable by evolution due to its evolutionary pathway to complexity. To overcome this bias and build a new, potentially better protein world we need to develop design principles that are not based on reinforcing the current set of “rules.” If we are successful in elucidating these new more rational design principles, we will potentially be able to build a new protein world and even new forms of life that can live longer and healthier than us. After all, we humans possess the power of systematic rational design as well as tools of directed and accelerated evolution [43,44,45,46]. But let us start by designing a new digital polymer.

Potential pitfalls and ways to overcome them. Even though the recognition elements in NP2 are placed with regular spacing (every 6 sigma bonds) on the peptide scaffold, the resulting minimum energy conformation of a double-stranded complex may not be as periodic as DNA double helix where a regular 3.4 Å spacing is ensured by π–π stacking. This potential lack of uniform periodicity, also known as Schrodinger’s aperiodic crystal requirement for information-bearing polymers (Schrodinger postulated that this requirement is needed for recognition by enzymes), is not necessarily a problem for many applications such as strand displacement. Strand displacement has already been demonstrated with α-helix based peptides and would benefit significantly from the acceleration that is expected for a more flexible single peptide chain compared to a rigid α-helix structure [47].

To address the recurrent concern about α-helix or β-sheet formation, I expect the hydrogen bonds of the code (R–D and G–G) to be more favorable than NH–OC hydrogen bonds as in NP1. As with NP1, this is expected because more compact interactions are entropically more favorable than the spread-out NH–OC interactions.

When planning synthesis, one should keep in mind that even at high coupling efficiency, solid-state reaction yields drop off rapidly, limiting the length of peptide or peptoid strands at sub-50-mer lengths (99% efficiency per-step gives less than 80% yield after 25 steps [48]). The high coupling efficiencies may be lower due to different monomers. However, the hybridization energy per base will likely be higher compared to DNA due to the absence of negative charges on single strands. This may enable shorter strands of NP2 be sufficient for building nanostructures and behaviors analogous to ones in structural and dynamic DNA nanotechnology.

Future outlook. A digital polymer made of only natural amino acids could be synthesized in vivo, enabling a multitude of biomedical applications for molecular programming—imagine being able to shape proteins into structures similar to DNA or RNA origami via co-translational folding, or being able to attach DNA-like strands in precise locations on protein surfaces to enable complex protein-based chemical reaction networks (CRNs) in vivo. Furthermore, biotechnological production of proteins is generally more scalable than that of nucleic acids. Perhaps one day we will grow designer nanoscale machines built with principles of DNA programming with the ease and cost of a cheese factory.

3 New Polymer 3 (NP3)

Design. DNA, NP1, and NP2 use hydrogen bonds in recognition elements. This puts upper bounds on the thermal stability of architectures designed with these polymers. NP3 aims to make a covalent recognition code for applications requiring very stable architectures. Is it possible to use stronger covalent bonds instead of weaker non-covalent ones to encode recognition between two strands of a digital polymer? Nature evolved to rely on weaker non-covalent interactions in many biological digital polymers, presumably to enable many dynamic behaviors at physiological temperatures. For example, the base pairing energy in DNA (hydrogen bonding + π–π stacking) is typically in 0.4–20.0 kJ/mol range; and more than 99.999% of proteins have free energy of folding below 33.5 kJ/mol [41]. Covalent bonds are an order of magnitude stronger, for example: N–H (386 kJ/mol), C–N (305 kJ/mol), C = C (602 kJ/mol), and C = N (615 kJ/mol). The main concern one may imagine with using covalent bonds for constructing a recognition code is that it would not allow error correction, i.e., a single incorrect bond will be so strong that it would prevent the system from reaching the desired global thermodynamic minimum. The strongest non-covalent DNA bonds are labile above 90 °C allowing for the global minima to be reached by slow annealing. Thermal annealing of covalently bound structures will likely destroy the constituent molecules. However, some covalent bonds can be made labile under mild conditions at room temperature, for example, in the presence of catalysts that lower their transformation activation energy. These bonds are known as “dynamic covalent.” I propose to use dynamic covalent bonds to construct NP3 (Fig. 4).

Fig. 4
A set of multiple reversible reactions of Imine formation, Boronate ester formation, Nitroaldol reaction, Hemiacetal formation, esterification, Olefin metathesis, disulfide exchange, and transesterification on the left. The molecular structure of N P 3 with various functional groups on the right.

Examples of dynamic covalent bonds (left and middle) and an example structure of New Polymer 3 duplex (right) made with ester and disulfide bonds from a self-complementary strand. Red Rs can potentially be any side chain

Full size image

Like NP2, NP3 uses only natural amino acids and relies on a 3-letter code constructed from heterophilic (S-D) and homophilic (C–C) interactions. Coincidentally, both ester and disulfide bonds can be rendered labile under the same conditions—low pH (ester) and reducing environment (S–S bond)—which is not the case for many other pairs of dynamic covalent bonds that may require incompatible conditions to become labile simultaneously [49].

Existing approaches to design a simple recognition code with dynamic covalent bonds. Dynamic covalent bonds have been explored extensively for the purposes of drug discovery, material design, and other goals [50]. Substantial efforts have been directed toward creating two-letter [51, 52] and even four-letter [53] heterophilic codes. The main challenge has been overcoming kinetically-trapped species with some recent successes for two-letter systems [54].

Potential pitfalls and ways to overcome them. In case the example shown in Fig. 4 does not form, there are many degrees of design freedom to explore. For example, artificial amino acids such as catechol and boronic acid can be incorporated [52]. Artificial amino acids would also allow positioning groups that form dynamic covalent bonds on aromatic residues such as benzene, which can enable stacking bonds for a more regular periodicity of duplex via π-π stacking like in DNA. When planning synthesis, consideration should be made for how non-canonical amino acids for the recognition elements of NP2 might be efficiently protected and deprotected in conventional solid-state peptide and peptoid syntheses. If protecting these groups is restrictive, post-polymerization reactions (i.e., thiolene click chemistry) to append desired recognition moieties can be explored. If one plans to use NP3 for in vivo applications, they should carefully plan the fate of the assemblies given potential swings in pH (e.g., acidic lysosomes), redox potential, etc.

4 Example Applications

The new digital polymers described above will enable many new capabilities unattainable with DNA. Some of the capabilities were summarized in Table 1. Below are a few more examples.

Activatable strand displacement. Overcoming leak reactions in strand displacement circuits is a significant challenge for dynamic DNA nanotechnology. One way to reduce or eliminate leak is to introduce a single covalent base pair such as cysteine-cysteine (C–C) from NP3 to a branch migration domain of a double-stranded complex based on non-covalent bonds (NP1 or NP2). The displacement will proceed only if the one special base is activated, e.g., with a reducing agent. The activation can be designed to be controlled by light [55] and other stimuli, expanding the complexity and versatility of dynamic behaviors that can be implemented with molecular programming.

Bio-orthogonality for in vivo molecular programming. As mentioned above, most nucleic acid analogs such as PNA were designed to bind nucleic acids. This limits their applications in vivo due to their interference with endogenous chemistry. NP2 and NP3 can be used to build a system for molecular programming in vivo that is orthogonal to existing nucleic acids.

Enzymeless translation and self-replication. The covalent code of NP3 makes binding by a single monomer stable. Simply by mixing a single template strand with free monomers, a new complementary strand can be templated. A ligation mechanism can be introduced (e.g., via terminal alkene methasesis) that would zip the new templated strand. After this, the complex can be “melted” via labilization of the covalent bonds, and the process repeated. Depending on whether the new monomers are the same as or different from monomers in the templating strand, this process can be conceptually viewed as replication or translation. Neither process would require enzymes, unlike their biological counterparts. Enzyme-free information transfer in digital polymers has been an focus of experimental efforts for many [56]. Enzyme-free protein replication (protein PCR) has been but the wild dream of few.

Molecular breadboard with 100 × pattern density of DNA origami. Using a version of NP1, it is also possible to design a fully addressable breadboard with average functionalization spacing of 0.6 nm (compare to current ~ 6 nm boards). I leave out the details of this design to stimulate the imagination of the reader.

5 Conclusions

Digital technologies are becoming essential to our existence. The digital paradigm enables construction of extremely complex hardware and software from simple building blocks. The original digital code can be found in life. The digital nature of naturally-evolved polymers such as DNA and proteins is essential to the existence and proliferation of life, by allowing robust information flow through precise encoding of genetic/structural information in sequences of base pairs/amino acids. While molecular programming has so far been dominated by work with DNA, chemists have been developing artificial polymers for many decades, though only a few of these constitute digital polymers, in which the sequence is precisely controlled down to single monomers. The few synthetic digital polymers that do exist have been shown to be superior to their less precise non-digital counterparts for applications in materials [57], molecular electronics [58], tracking [10], and molecular programming, because they allow precise control through their sequence of melting point, bandgap energy, compact mass-spec signature, and sequence-dependent recognition, respectively. Yet, despite this progress, these non-biological digital polymers are still largely underutilized.

DNA will likely remain the molecule of choice for molecular programming for the foreseeable future due to its current practical advantages. But I hope that the ideas outlined here will encourage exploration of new digital polymers with expanded capabilities—both experimentally by chemists and theoretically by molecular programmers. Nature initially used RNA and DNA as the substrate for life during the “RNA world” before evolving construction based on proteins, a paradigm shift that enabled much more diverse forms and functions. Similarly, the approaches outlined here could lead to more diverse nanotechnologies built with molecular programming.