{"id":676782,"date":"2020-07-23T12:37:41","date_gmt":"2020-07-23T19:37:41","guid":{"rendered":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/?p=676782"},"modified":"2020-10-06T15:55:37","modified_gmt":"2020-10-06T22:55:37","slug":"researchers-use-a-strand-displacing-dna-polymerase-to-do-biocomputing","status":"publish","type":"post","link":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/blog\/researchers-use-a-strand-displacing-dna-polymerase-to-do-biocomputing\/","title":{"rendered":"Researchers use a strand-displacing DNA polymerase to do biocomputing"},"content":{"rendered":"\n
\"Animated<\/figure>\n\n\n\n

All around us, biochemical systems are regulating a variety of natural processes, from our body\u2019s ability to protect our skin through the timely production of melanin to plants\u2019 ability to convert carbon dioxide into carbohydrates and oxygen using sunlight. Replicating and programming such complex systems\u2014essentially creating computing networks capable of operating in biological environments\u2014offers a unique opportunity to go where traditional silicon-based computers can\u2019t. For example, with synthetic biocompatible controllers, we\u2019re looking at the potential for targeted medical therapies. Think cancer treatments that attack only dangerous cells, sparing healthy ones, or capsules that delivery drugs or antibodies at opportune times.<\/p>\n\n\n\n

Building these systems requires programming chemicals, and synthetic DNA is an ideal raw material to work with. Highly programmable, like transistors in silicon technology, and biocompatible, DNA can be used to implement chemical reaction networks (CRNs), a programming language for representing chemistry and biological processes, or the algorithmic and logical functions in traditional computing terms. Existing architectures implementing CRNs via DNA come in two varieties: DNA-only systems and multienzyme DNA systems. Researchers have been using DNA-only systems to build interesting digital logic circuits, such as circuits that can compute the square root of four-bit numbers (opens in new tab)<\/span><\/a> and others that can classify handwritten digits (opens in new tab)<\/span><\/a> by implementing pretrained neural networks. But both DNA-only and multienzyme DNA architectures have drawbacks, mainly slow rates and leaky reactions for DNA-only systems and increased biological complexity that can restrict environmental conditions for multienzyme systems. These challenges can limit the size of the systems, as well as their introduction into biological environments.<\/p>\n\n\n\n\t

\n\t\t\n\n\t\t

\n\t\tPODCAST SERIES<\/span>\n\t<\/p>\n\t\n\t

\n\t\t\t\t\t\t
\n\t\t\t\t\n\t\t\t\t\t\"Illustrated\n\t\t\t\t<\/a>\n\t\t\t<\/div>\n\t\t\t\n\t\t\t
\n\n\t\t\t\t\t\t\t\t\t

AI Testing and Evaluation: Learnings from Science and Industry<\/h2>\n\t\t\t\t\n\t\t\t\t\t\t\t\t

Discover how Microsoft is learning from other domains to advance evaluation and testing as a pillar of AI governance.<\/p>\n\t\t\t\t\n\t\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\n\t\t\t\t\t\t\tListen now\t\t\t\t\t\t<\/a>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t<\/div>\n\t<\/div>\n\t\n\n\n

We propose a promising solution with a novel method for implementing chemical reaction networks that incorporates one enzyme. The idea is to replace the most fundamental unit of DNA computing, namely, toehold mediated strand displacement (TMSD), a DNA-only architecture, with polymerase-based strand displacement (PSD). Using DNA polymerase-based systems has several benefits: The polymerase enzyme gives an external energy source to the system, which is usually required; can synthesize new DNA strands, unlike enzyme-free systems; and can be potentially really fast. We present polymerase-based strand displacement in the paper \u201cUsing Strand Displacing Polymerase to Program Chemical Reaction Networks<\/a>,\u201d which was published in the Journal of the American Chemical Society.<\/p>\n\n\n\n

Toehold mediated strand displacement vs. polymerase-based strand displacement<\/h3>\n\n\n

In 2006, researchers from the California Institute of Technology introduced enzyme-free logic computing (opens in new tab)<\/span><\/a>, which has since become the go-to approach for programming chemical reaction networks among computer scientists. Because of its simplicity and tunability, TMSD in particular has become one of the most fundamental DNA computing architectures. TMSD and PSD are similar in design, though. But first, a few things about DNA.<\/p>\n

DNA strands are composed of sequences of the chemical bases adenine (A), guanine (G), cytosine (C), and thymine (T); DNA strands combine to form complexes, with A joining to its complementary base T and C to its complementary base G. Usually, an abstract representation of DNA strands is used to make it easier to design complicated architectures, which can have multiple DNA strands, each with sequences numbering in the tens or hundreds. So, for example, a portion of a single-stranded DNA with the sequence TACGTATGAATCAG might be referred to as domain<\/em> o <\/em>(Figure 1a). The reverse complement of that sequence would be ATGCATACTTAGTC, or more simply domain o*<\/em>. This way, we don\u2019t need to worry about actual sequences but can operate abstractly at the domain level. Sequences on a DNA strand can also be split into different parts, each with a separate domain name. In the domain o<\/em> example, the sequence could be split after the first G: TACG TATGAATCAG. The first portion of the sequence could be domain t<\/em> and the second domain o<\/em>; the overall strand is t<\/em> o.<\/p>\n\n\n

\"A
Figure 1: Highly programmable and biocompatible, synthetic DNA is an ideal raw material for implementing chemical reaction networks. In Section (a) in the above figure, DNA is represented in several ways: as a double helix; as a sequence of its chemical bases joined with a complementary sequence; and as a more abstract representation in which the letters o <\/em>and o* <\/em>stand in for the specific sequences. To make it easier to design complicated architectures, the researchers use the latter representation. Sections (b) and (c) illustrate toehold mediated strand displacement (TMSD) and polymerase-based strand displacement (PSD), respectively. Polymerase-based strand displacement incorporates one enzyme, polymerase, represented by the red oval.<\/figcaption><\/figure><\/div>\n\n\n\n

In TMSD, an input DNA strand consisting of complementary domains\u2014let\u2019s call the strand t*<\/em> o*<\/em>\u2014binds with an exposed single-stranded, or toehold<\/em>, portion of a double-stranded DNA complex t<\/em> o <\/em>(Figure 1b). This binding displaces the output strand in a \u201ctug of war\u201d between the duplicate portion of the input strand\u2014the o*<\/em> domain\u2014and the o*<\/em> domain that already exists in the complex. In PSD, a shorter input DNA strand without the o*<\/em> domain attaches to the exposed portion of the complex; once there\u2014and in our design, it\u2019s a permanent bind\u2014the polymerase enzyme elongates the input strand by printing new DNA, a t* o*<\/em> copy, displacing the output in the double-stranded complex (Figure 1c).<\/p>\n\n\n\n

Though similar in design, the addition of just one enzyme, polymerase, offers promising benefits. TMSD challenges include leaky reactions, which require workarounds that result in slower computation and\/or greater system complexity. One reason for leaks is the presence of the output domain in the input, creating an overlap between the signals that increases the potential for error. In PSD, there is no overlap. The fact that there\u2019s no longer a tug of war between strands, a time-consuming process, also contributes to faster computation times. Additionally, polymerase provides an external energy source, which allows for more complex computation within a reasonable time frame. For example, in 24 hours, DNA-only architectures may only be able to complete three layers of circuits; the faster computation offered by PSD can potentially complete 20 layers in the same period. This external energy source comes from the hydrolysis of the A, G, C, and T bases and the sugar and phosphate molecules that accompany them, collectively known as nucleoside triphosphates (NTPs). For example, it takes DNA-only architectures 10 hours to compute the square root function of a four-bit input; the faster computation offered by PSD can complete the same function within 40 minutes (opens in new tab)<\/span><\/a>. With PSD, since external NTPs and polymerase are used, several hundreds of bases can be displaced even if the toehold is short.
<\/p>\n\n\n\n

\"\"<\/a>
Figure 2: In chemistry, unimolecular and bimolecular CRNs are considered the two most fundamental gates. On the left are a typical unimolecular reaction and a typical bimolecular reaction. A unimolecular reaction has one input (A) and a bimolecular reaction has two inputs (A and B). Their domain-level DNA implementation using the PSD architecture is shown on the right. For both reactions, it\u2019s a two-step process. Note that for the bimolecular reaction, input B is a double-stranded gate(Gbi), and because it is required to be in a double-stranded form or a single-stranded form, depending on the system, there is a Linker reaction to facilitate the conversion (for more information, see theoretical paper).<\/figcaption><\/figure><\/div>\n\n\n\n

Designing unimolecular and bimolecular reactions<\/h3>\n\n\n\n

Much like logic gate AND and logic gate OR are considered universal gates in traditional computation, unimolecular and bimolecular CRNs are considered the two most fundamental gates in chemistry as far as design and implementation. In unimolecular reactions, only one reactant is used; in a bimolecular reaction, two reactants are used. Depending on the design, both CRNs are capable of producing either a single output or multiple outputs. Complex CRNs with more than two inputs and\/or outputs can be built using these basic CRNs, enabling more advanced processes.<\/p>\n\n\n\n

In theory, as described in the earlier paper \u201cImplementing Arbitrary CRNs Using Strand Displacing Polymerase,\u201d (opens in new tab)<\/span><\/a> we demonstrate that we can implement unimolecular and bimolecular CRNs using PSD and do any arbitrary computations. We also apply these CRNs in an autocatalytic amplifier, a molecular-scale consensus network, and a dynamic rock-paper-scissor oscillatory system. In our architecture, unimolecular reactions and bimolecular reactions are both implemented in a two-step process. In the unimolecular reaction, A combines with an auxiliary gate to release intermediate strand I, which then combines with the next gate to produce output strand B. In the bimolecular reaction, input A combines with input B to produce intermediate strand I, which combines with the next gate to produce output strand C.<\/p>\n\n\n\n

\"\"
Figure 3: The main experimental result of our work is a catalytic amplifier. The amplifier workflow is cyclic: Input strand A performs a two-step reaction and gets recycled back to the solution along with the output release. The recycled input continues the reaction with other gates while the output attaches with a fluorescent complex for recording its activity. The graph on the right shows the experimentally observed catalytic activity. It demonstrates that when input quantity is less than 100 percent, the output still gets fully triggered. We tried inputs from 0 percent to 100 percent (in the graph, 0.00x = 0 percent, 4.00x = 40 percent, 6.00x = 60 percent, etc.), as compared to all the supporting gates, which are at 100 percent.<\/figcaption><\/figure>\n\n\n\n

Implementing PSD in lab<\/h3>\n\n\n\n

In our Journal of the American Chemical Society<\/em> paper<\/a>, we take the first step toward demonstrating in lab that our architecture can implement CRNs, showing the fundamentals of PSD systems and designing an in vitro<\/em> catalytic amplifier. A catalytic amplifier is a unimolecular reaction with two outputs, but one of the outputs is also input. A catalytic amplifier is a strong proof of concept. Not only is it a show of complex DNA design, but it\u2019s also sustainable.<\/p>\n\n\n\n

Unlike in noncatalytic systems, where you need the same amount of input to trigger a certain amount of output, in a catalytic system like a catalytic amplifier, you can trigger 100 percent output with, let\u2019s say, only 10 percent input. That\u2019s because of the presence of a catalyst, or a fuel, which helps recycle the input so it can be used again to release more output. Even if there is a signal loss, you can still get experimentally full output activation from lower levels of input. For example, a catalytic amplifier could be used for fast signal restoration in deep circuits and neural nets.<\/p>\n\n\n\n

In executing a catalytic amplifier, as well as other fundamental reactions, we show we\u2019re able to tune two defining properties of a chemical reaction: stoichiometry and reaction rate. Our PSD architecture will only activate the same amount of output as input, allowing us to control the amount of output via the amount of input (stoichiometry), and the speed at which output is produced can be controlled by varying the lengths of the input strands.<\/p>\n\n\n\n

Infinite possibilities<\/h3>\n\n\n\n

This was the first study on implementing CRNs using PSD, so for us, the goal was to explore the fundamentals. Our work mainly focuses on the design and demonstration of the basic properties of such systems, such as tuning the reaction speed and controlling the stoichiometry, and with the catalytic amplifier, we put the fundamentals to work. Next steps include in vitro demonstrations of larger-scale autocatalytic systems such as oscillators, linear controllers, and pulses, which are the accepted gold standards in the field. Such complex biochemical controllers can have computing applications in the biological context, where traditional silicon can\u2019t reach, offering new possibilities in medical care, as mentioned above, agriculture, energy, molecular biology, computing, and sensor networks, among other areas.<\/p>\n\n\n\n

DNA computing is decades behind its silicon counterpart, but we see fast growth as a real possibility, with the field learning from current technology. And thanks to design software like Visual DSD<\/a> from Microsoft Research, DNA computing will be more streamlined, and designing and testing a new architecture\u2014without leaks\u2014should be easier, opening the way for some very cool applications.<\/p>\n\n\n\n

Acknowledgment: This work was conducted by Shalin Shah (opens in new tab)<\/span><\/a>, Jasmine Wee (opens in new tab)<\/span><\/a>, Tianqi Song (opens in new tab)<\/span><\/a>, Luis Ceze (opens in new tab)<\/span><\/a>, and Karin Strauss<\/a>, jointly led by John Reif (opens in new tab)<\/span><\/a> at Duke University and Yuan-Jyue Chen<\/a> at Microsoft Research. Shah was a Microsoft Research intern at the time of the work.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

All around us, biochemical systems are regulating a variety of natural processes, from our body\u2019s ability to protect our skin through the timely production of melanin to plants\u2019 ability to convert carbon dioxide into carbohydrates and oxygen using sunlight. Replicating and programming such complex systems\u2014essentially creating computing networks capable of operating in biological environments\u2014offers a […]<\/p>\n","protected":false},"author":38838,"featured_media":676827,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":null,"msr_hide_image_in_river":0,"footnotes":""},"categories":[1],"tags":[],"research-area":[13553],"msr-region":[],"msr-event-type":[],"msr-locale":[268875],"msr-post-option":[],"msr-impact-theme":[],"msr-promo-type":[],"msr-podcast-series":[],"class_list":["post-676782","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-research-blog","msr-research-area-medical-health-genomics","msr-locale-en_us"],"msr_event_details":{"start":"","end":"","location":""},"podcast_url":"","podcast_episode":"","msr_research_lab":[],"msr_impact_theme":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-academic-programs":[],"related-groups":[],"related-projects":[],"related-events":[],"related-researchers":[{"type":"guest","value":"shalin-shah","user_id":"600375","display_name":"Shalin Shah","author_link":"Shalin Shah<\/a>","is_active":true,"last_first":"Shah, Shalin","people_section":0,"alias":"shalin-shah"},{"type":"user_nicename","value":"Yuan-Jyue Chen","user_id":35057,"display_name":"Yuan-Jyue Chen","author_link":"Yuan-Jyue Chen<\/a>","is_active":false,"last_first":"Chen, Yuan-Jyue","people_section":0,"alias":"yuanjc"}],"msr_type":"Post","featured_image_thumbnail":"\"\"","byline":"Shalin Shah<\/a> and Yuan-Jyue Chen<\/a>","formattedDate":"July 23, 2020","formattedExcerpt":"All around us, biochemical systems are regulating a variety of natural processes, from our body\u2019s ability to protect our skin through the timely production of melanin to plants\u2019 ability to convert carbon dioxide into carbohydrates and oxygen using sunlight. Replicating and programming such complex systems\u2014essentially…","locale":{"slug":"en_us","name":"English","native":"","english":"English"},"_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/posts\/676782","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/users\/38838"}],"replies":[{"embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/comments?post=676782"}],"version-history":[{"count":16,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/posts\/676782\/revisions"}],"predecessor-version":[{"id":696444,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/posts\/676782\/revisions\/696444"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media\/676827"}],"wp:attachment":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media?parent=676782"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/categories?post=676782"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/tags?post=676782"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=676782"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=676782"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=676782"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=676782"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=676782"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=676782"},{"taxonomy":"msr-promo-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-promo-type?post=676782"},{"taxonomy":"msr-podcast-series","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-podcast-series?post=676782"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}