Method for determining resolution of RNA heterogeneity in three-dimensional structure of RNA

Abstract

The invention discloses a method for determining RNA heterogeneity in a three-dimensional structure of RNA, which comprises the following steps: S3, purifying folded RNA, freezing, preparing a sample, vitrifying, and evaluating under a freezing electron microscope to obtain a carrier net with optimal ice thickness and uniform particle size and distribution; s4, the grid carries out small-amount data collection, data processing is paused after two-dimensional (2D) classification is completed so as to carry out preliminary evaluation, if the preliminary evaluation has obvious RNA characteristics, large-scale cryoelectron microscope data collection and subsequent data processing are carried out so as to solve RNA heterogeneity, and if the preliminary evaluation does not have obvious RNA characteristics, large-scale cryoelectron microscope data collection and subsequent data processing are carried out so as to solve RNA heterogeneity; if not, trying different RNA preparation and / or vitrification conditions to obtain an optimized RNA sample, carrying out small-amount data collection again until obvious RNA characteristics exist, carrying out large-scale cryoelectron microscope data collection, and then carrying out data processing to solve RNA heterogeneity; and S5, most RNA experiences conformational changes when executing functions, and the conformational changes are separated more effectively by combining a standard cryoelectron microscope data processing flow and a three-dimensional classification algorithm. The present invention assists in implementing cryo-EM structure assays, assisted by a cryo-EM data processing flow, to parse RNA dynamics and conformational variations, and an RNA modeling algorithm to generate atomic coordinates based on a medium to high resolution cryo-EM density map.

Description

Technical Field

[0001] The present invention relates to the construction of a three-dimensional structure model of RNA, and in particular to a method for solving RNA heterogeneity in determining the three-dimensional structure of RNA, a method for determining the three-dimensional structure of RNA, and an application of the method for determining the three-dimensional structure of RNA. Background Art

[0002] RNA participates in various basic biological processes, such as catalysis, transcriptional regulation, translation, and signaling pathway regulation. The three-dimensional structure of RNA mainly determines their functions, for example, ribozymes form a compact core structure to catalyze reactions; riboswitches regulate gene expression by sensing different ligands and environmental changes; long noncoding RNA (lncRNA) folds into specific structures and interacts with proteins to perform its functions.

[0003] When performing functions in vitro or in vivo, structured RNAs may undergo large conformational changes, and individual nucleotides may also undergo local dynamic changes during the execution of their functions. This complicates the preparation of homogenous RNA samples for structural determination. In addition, traditional structural biology techniques, such as X-ray crystallography, nuclear magnetic resonance (NMR), and cryo-electron microscopy single particle analysis (Cryo-EM SPA), have their technical limitations when dealing with the heterogeneity of different RNAs, which together pose a great challenge to the high-resolution three-dimensional structural analysis of RNA. X-ray crystallography usually requires a large amount of RNA sample, and the overall structural heterogeneity of RNA should be as small as possible to facilitate the stacking and lattice formation of molecules during crystal growth. Conformational changes of structural domains can sometimes be observed. NMR is mainly used to study the interconversion of individual nucleotides and bases in small RNAs (average size of 30 nucleotides (30-nt)) in low-energy steady states and high-energy excited states. Cryo-EM SPA can resolve the global conformational changes of larger RNAs and their complexes and the local dynamics of nucleotides and ions at near-atomic resolution. However, the smallest RNA molecule studied so far is 88 nt, and most RNA cryo-EM structures have been solved at resolutions of medium To low resolution The scope of these RNAs is limited.

[0004] Although existing methods have made RNA structure determination more efficient, RNA structures are still underrepresented in the protein data bank (PDB) and the electron microscopy data bank (EMDB) compared to proteins, which may be one of the main reasons why the accuracy of RNA structure prediction is lower than that of proteins. This highlights the need for a systematic approach and process that can be widely used for cryo-EM structure determination of RNA with clear functions. Even for the more simplified protein cryo-EM structure determination, many variables in sample preparation and vitrification need to be iteratively optimized, often requiring the operator's expertise and experience. This makes existing methods difficult to standardize because their effects vary from sample to sample.

[0005] The above background technology is for facilitating the understanding of the present invention and is not a known technology disclosed to the general public before the application of the present invention. Summary of the invention

[0006] In view of the above-mentioned defects, the present invention provides a method for in vitro transcription and purification of RNA and RNA folding in determining the three-dimensional structure of RNA, which aims to improve at least one of the problems mentioned in the background technology.

[0007] The method of the present invention is developed for RNAs that perform functions by forming three-dimensional structures. Structural analysis of these RNAs with well-defined functions can elucidate their molecular mechanisms. The method of the present invention determines the cryo-EM structure of RNAs of different functions and molecular weights, covering a wide range of conditions.

[0008] The method of the present invention, "Ribosolve", provides a rapid cryo-EM-guided RNA structure determination process, as well as other methods such as oligomers and higher-order assemblies to facilitate cryo-EM analysis, and the use of large scaffold RNA to achieve cryo-EM structure determination of small RNAs.

[0009] The technical solution is: a method for solving RNA heterogeneity in determining the three-dimensional structure of RNA, comprising the following steps:

[0010] S3, the purified folded RNA is then cryo-vitrified and evaluated under cryo-EM to obtain a grid with optimal ice thickness and uniform particle size and distribution;

[0011] S4, a small amount of data is collected on the grid. Data processing is paused after two-dimensional (2D) classification is completed for preliminary evaluation. If there are obvious RNA features in the preliminary evaluation, large-scale cryo-EM data collection and subsequent data processing are performed to resolve RNA heterogeneity. If there are no obvious RNA features, different RNA preparation and / or vitrification conditions need to be tried to obtain optimized RNA samples and small amounts of data are collected again until there are obvious RNA features. Large-scale cryo-EM data collection is then performed before data processing is performed to resolve RNA heterogeneity.

[0012] S5, Most RNAs undergo conformational changes when performing their functions. The standard cryo-EM data processing pipeline combined with the 3D classification algorithm can more effectively separate conformational changes.

[0013] The present invention also provides a method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA.

[0014] The technical solution is: a method for solving RNA heterogeneity in determining the three-dimensional structure of RNA, comprising the following steps:

[0015] S3, RNA sample vitrification, cryo-EM observation, and data collection;

[0016] S4, Data processing to address RNA heterogeneity.

[0017] Further, S3 includes the following steps:

[0018] S19, prepare the required amount of RNA samples under optimal conditions for vitrification and cryo-electron microscopy analysis;

[0019] S20, the temperature and relative humidity of the Mark IV Vitrobot were set to 4 °C and 100%, and filter papers were assembled in the chamber and cryogen container for vitrification;

[0020] S21, select the appropriate type of grid and perform glow discharge, wherein the copper grid is mainly used in the screening stage, and the support membrane is used when the final concentration is lower than 1 μM and the RNA is greater than 100 kDa;

[0021] S22, load the grid onto the Vitrobot, add 2-3 μL of sample onto the grid, use filter paper to absorb the excess sample and inject cryogen, and transfer the grid to the grid box pre-cooled with liquid nitrogen;

[0022] S23, load the automated grid into a pre-cooled box, transfer the box into a NanoCab filled with liquid nitrogen, and then dock the box onto the microscope to check if the microscope is ready for imaging;

[0023] S24, open the EPU and select Atlas mode to get an overview of the ice distribution on the grid; move the sample stage to the area of ​​interest and acquire images in GridSquare mode; find the concentric height, move the sample stage to the hole with the appropriate ice layer, and enter Hole / EucentricHeight mode; collect cryo-EM images in Data Acquisition mode;

[0024] S25, if the carrier grid is not ideal and is not suitable for data collection, remove the carrier grid and repeat S24 to screen the next carrier grid;

[0025] S26, after completing the screening of all automatic grids, if the ice thickness is not ideal, repeat S20-S25 to screen the grid, glow discharge time and Vitrobot parameters until the optimal ice thickness is obtained; if the RNA sample state is not suitable for data collection, screen the variables in in vitro transcription and RNA folding until the best micrograph suitable for data collection is observed;

[0026] S27, activate the “ice filter” in the EPU to select holes based on ice thickness determined by image intensity;

[0027] S28, Set up automatic data collection according to the EPU protocol and check the microscope and camera settings to ensure that data collection proceeds smoothly.

[0028] Further, S4 includes the following steps:

[0029] S29, import all cryo-EM images and use MotionCor2 to perform "MotionCorrection", set "Number of patches X,Y" to "5,5", and set "Other MotionCor2 parameters" to "-Iter 10", the dose per frame is provided accordingly, and the rest of the parameters remain default;

[0030] S30, perform "Contrast transfer function estimation" using CTFFIND4 with default settings, but set "Amount ofastigmatism" to Contrast ring out of The micrographs were visually inspected and selected;

[0031] S31, using Blob picker in cryoSPARC, reference-based template matching in Relion, and reference-based CNN particle picking in EMAN2;

[0032] S32, using the imported micrograph in S30 and the "input coordinates" generated in S31 to perform "particle extraction", setting an appropriate "particle frame size";

[0033] S33, perform "2D classification" on the scaled particles, set "Number of classes" to 200, "Use EM algorithm" to "Yes", "Number of EM iterations" to 50, "Use VDAM algorithm" to "No", and "Mask diameter" to and set Limit resolution E-step to The rest of the parameters remain default;

[0034] S34, visually inspect the classification results in “Subset selection” and discard those unwanted RNA particles and classes without obvious RNA features;

[0035] S35, repeat S33-S34 2-3 times until most of the unnecessary particles are discarded; if no RNA feature is observed, terminate the data processing until the 2D classification results reveal the RNA feature and then perform data processing;

[0036] S36, import particles from S35 and generate a "3D initial model", set the "Number of classes" to 6, and keep the other parameters as default;

[0037] In S37, perform “3D classification” by importing all particles in S35 and a “reference image” in S36, set “Number of classes” to 6, and set “Limit resolution E-step” to The rest of the parameters remain default;

[0038] S38, repeating S34 to select particles from the desired category, and then repeating S32 to extract the particles without scaling;

[0039] S39, performing preliminary "3D automatic refinement" by importing the selected particles in S38 and a "reference image" in S37;

[0040] S40, perform spectrum sharpening in "Post-processing" and import the "half unfiltered spectrum" and "solvent mask" in S39;

[0041] S41, cryo-EM map visualized in Chimera;

[0042] S42, repeat S37-S41 to obtain a better resolution of the flexible region density;

[0043] S43, repeating steps S39-S41 for the selected category of particles and providing a corresponding "reference mask" to obtain a final structure that better resolves RNA heterogeneity;

[0044] S44, import "CTF refinement", "grains" and "post-process STAR files" from S43, first "estimate magnification";

[0045] S45, import "particles" from S44 into "CTF refinement", perform "CTF parameter fitting", set "Fit defocus" and "Fit B factor" to "per particle", "Fit astigmatism" to "per micrograph", "Esti mate beamtilt" and "Also estimate trefoil" to "yes;

[0046] S46, import “micrographs” from S29, “particles” from S44, and “post-processed STAR files” from S43 into “Bayesian refinement”, set “train optimal parameters” to “yes”, set “perform particle refinement” to “no”, and keep other parameters as default;

[0047] S47, in "Bayesian Refinement", import the "Optimization Parameters File" from S46, set "Train Best Parameters" to "No", set "Perform Granular Refinement" to "Yes", and keep other parameters as default;

[0048] S48, repeating S39-S41 to obtain a final structure from the refined particles;

[0049] S49, analyze the "local resolution" of the final map by providing the required input files and the "user-provided B-factor" from S48, as well as the corresponding "calibrated pixel size", set "use relion" to "yes", and import the "detector MTF (STAR ​​file) detector MTF" from the corresponding detector manufacturer;

[0050] S50, Local resolution density map of S49 colored for presentation in Chimera.

[0051] Further, the use of Blob picker in cryoSPARC includes the following steps:

[0052] S311, start a new "Blob Picker" job and import the micrographs in S30;

[0053] S312, check “Show advanced” and set the expected minimum and maximum particle diameters. For Tetrahymena group I introns, this value is set to 50 and

[0054] S313, open "Circular blob" and run the job, keeping the rest of the parameters as default settings;

[0055] S314, after completing the "Blob Picker" job, start a new "Inspect Particle Picks" job. Import the micrographs and particles from the "Blob Picker", keep the parameters as default, and adjust the "PowerHistogram" to visually inspect the particle picking results for each micrograph;

[0056] S315, when the selected particles are as many as possible and uniform, click "Done Picking".

[0057] Further, the use of reference-based template matching in Relion includes the following steps:

[0058] S311, import the micrographs from S30 and set the corresponding “Pixel size in micrographs”;

[0059] S312, import the results of the two-dimensional classification average, and set the "Lowpass filter references" to Set the Pixel size in references as required.

[0060] Further, the use of reference-based CNN particle picking in EMAN2 includes the following steps:

[0061] S311, navigate to "Particles->Interactive Particle Picking" in the GUI and import the micrograph from S30; interactively enter the "boxsize" and "ptclsize" values ​​determined from the micrograph. For the Tetrahymena group I intron, these values ​​are set to 256 and 200. Fill in the remaining parameters;

[0062] S312, manually select 15-20 reference images as good and background references;

[0063] S313, train the CNN model with the “Threshold1” value set to 2;

[0064] S314, click “Autobox” on a single micrograph to check the particle picking results;

[0065] S315, when the selected particles are as many as possible and uniform, click "Autobox All".

[0066] Furthermore, in S39, “Reference mask” is imported and “Use solvent-flattened FSCs” is set to “Yes” to improve the reconstruction effect of low molecular weight and low signal-to-noise ratio particles.

[0067] The present invention also provides a method for determining the three-dimensional structure of RNA.

[0068] The technical solution is: a method for determining the three-dimensional structure of RNA, the method comprising:

[0069] RNA in vitro transcription and purification and RNA folding methods;

[0070] The above-mentioned method for resolving RNA heterogeneity; and

[0071] RNA modeling and methods for manual revision and new modeling based on RNA structure prediction.

[0072] The present invention also provides an application of the method for determining the three-dimensional structure of RNA.

[0073] The technical solution is: an application of the above method for determining the three-dimensional structure of RNA to RNA of 70-nt or larger that forms a three-dimensional structure.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] In order to study the structure-function relationship of RNA at different resolutions, the present invention establishes an atomic model of RNA based on a cryo-electron microscopy density map.

[0076] The method of the present invention summarizes a class of universal methods for determining RNA tertiary structure, which may answer more biological questions and contribute to the continuous development of the field.

[0077] The method of the present invention allows iteration in RNA preparation conditions, including pH, buffer composition, ion concentration and additives (such as detergents). These conditions can be evaluated by the most commonly used PAGE and combined with other alternative methods that provide additional information on sample properties, such as dynamic lighter scattering (DLS and nanoDLS). We also allow iteration in RNA sample preparation (sample vitrification) conditions, such as different types of grids, support membranes, grid pretreatment and freezing conditions, which can be tested by cryo-EM sample screening and preliminary data collection. The optimized samples can be subjected to large-scale data collection and processing to resolve RNA dynamic heterogeneity at the best resolution, and use RNA modeling algorithms to generate atomic coordinates based on cryo-EM density maps, reaching near-atomic resolution.

[0078] This paper describes the rigorous screening and iterative optimization of RNA preparation conditions to enable cryo-EM structure determination, supplemented by a cryo-EM data processing pipeline to resolve RNA dynamics and conformational changes, and an RNA modeling algorithm to generate atomic coordinates based on moderate to high-resolution cryo-EM density maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 is a process flow chart of steps 01-26 of the present invention;

[0080] Figure 2 It is the process flow chart of steps 27-81 of the present invention. DETAILED DESCRIPTION

[0081] As used herein:

[0082] "Prepared from" is synonymous with "comprising." As used herein, the terms "comprising," "including," "having," "containing," or any other variation thereof, are intended to cover a non-exclusive inclusion. For example, a composition, process, method, article, or apparatus that comprises the listed elements is not necessarily limited to only those elements but may include other elements not expressly listed or inherent to such composition, process, method, article, or apparatus.

[0083] The conjunction "consisting of excludes any unspecified element, step, or component. If used in a claim, this phrase renders the claim closed-ended so that it does not include materials other than those described, except for conventional impurities associated therewith. When the phrase "consisting of" appears in a clause of the body of a claim rather than immediately following the subject matter, it limits only the elements described in that clause; other elements are not excluded from the claim as a whole.

[0084] When an amount, concentration, or other value or parameter is expressed as a range, a preferred range, or a range defined by a series of upper preferred values ​​and lower preferred values, this should be understood as specifically disclosing all ranges formed by any pairing of any range upper limit or preferred value with any range lower limit or preferred value, regardless of whether the range is disclosed separately. For example, when a range of "1 to 5" is disclosed, the described range should be interpreted as including ranges "1 to 4", "1 to 3", "1 to 2", "1 to 2 and 4 to 5", "1 to 3 and 5", etc. When a numerical range is described in this article, unless otherwise stated, the range is intended to include its end values ​​and all integers and fractions within the range.

[0085] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] Those skilled in the art will appreciate that the following examples are only used to illustrate the present application and should not be considered as limiting the scope of the present application. If specific conditions are not specified in the examples, they are carried out according to normal conditions or conditions recommended by the manufacturer. If the manufacturer is not specified for the reagents or instruments used, they are all conventional products that can be purchased commercially.

[0087] In these examples, parts and percentages are by mass unless otherwise indicated.

[0088] "Parts by mass" refers to the basic unit of measurement for expressing the mass ratio of multiple components. 1 part can represent any unit mass, such as 1g or 2.689g. If we say that the mass of component A is a parts and the mass of component B is b parts, it means the ratio of the mass of component A to the mass of component B is a:b. Alternatively, it means that the mass of component A is aK and the mass of component B is bK (K is an arbitrary number, indicating a multiple factor). It should not be misunderstood that, unlike the mass parts, the sum of the mass of all components is not limited to 100 parts.

[0089] "And / or" is used to indicate that one or both of the stated situations may occur, for example, A and / or B includes (A and B) and (A or B).

[0090] A method to address RNA heterogeneity, comprising:

[0091] S3, the purified folded RNA is then vitrified and evaluated under cryo-EM to obtain a grid with optimal ice thickness and uniform particle size and distribution.

[0092] S4, a small amount of data is collected on the grid. Data processing is paused after two-dimensional (2D) classification is completed for preliminary evaluation. If the preliminary evaluation has obvious RNA features, large-scale cryo-EM data collection and subsequent data processing will be carried out to resolve RNA heterogeneity. If there are no obvious RNA features, different RNA preparation and / or vitrification conditions need to be tried to obtain optimized RNA samples and small amounts of data collection is carried out again until there are obvious RNA features. Large-scale cryo-EM data collection is then carried out before data processing is carried out to resolve RNA heterogeneity.

[0093] S5, Most RNAs undergo conformational changes when performing their functions. The standard cryo-EM data processing pipeline combined with the 3D classification algorithm can more effectively separate conformational changes.

[0094] Example 1 Obtaining the secondary structure of RNA. The secondary structure finally obtained will be used in S61 or S70 for the prediction of RNA three-dimensional model

[0095] The accurate secondary structure of RNA can be obtained by existing methods, which can be obtained by high-throughput analysis using dimethyl sulfate (DMS), selective 2'-hydroxyacetylation by primer extension analysis (SHAPE), or by mutation spectra, cross-linking, and experimental or computational methods to detect single base pairing, structural proximity, and alternative RNA structures. No special regulations are made here, as long as RNA with a clear secondary structure and accurate RNA secondary structure information are obtained.

[0096] The purpose of obtaining RNA with a clear secondary structure and accurate RNA secondary structure information is: (i) RNA with a clear secondary structure is more likely to fold into a rigid and uniform three-dimensional structure; (ii) accurate secondary structure information is helpful for the interpretation of cryo-EM density maps and model construction, especially in the construction of medium-resolution cryo-EM models.

[0097] Example 2 RNA in vitro transcription and purification and RNA folding

[0098] An RNA in vitro transcription and purification and RNA folding method comprises the following steps:

[0099] S1, RNA in vitro transcription and purification, including the following steps:

[0100] S01, linearized template preparation, obtain the linearized template through polymerase chain reaction (PCR) or enzymatic digestion reaction.

[0101] Among them, polymerase chain reaction methods include:

[0102] S011, 400 ng of plasmid, 500 μL of 2× Phanta Max Master Mix, and 40 μL of forward and reverse primers were mixed in DEPC-treated water to a total volume of 1 mL.

[0103] The DNA polymerase and 2'-O-methylated reverse primer in this step can improve the uniformity of the 3' end.

[0104] S012, PCR reaction was performed according to the manufacturer's instructions.

[0105] S013, mix 1 μL of PCR product with 1 μL of 10× DNA loading buffer, load 1 μL onto a 1% (wt / vol) native agarose gel containing 1:10,000 (vol / vol) Ultra GelRed solution (use 5 μL of Ultra GelRed for 50 mL of agarose gel), and add 1 μL of D2000 marker.

[0106] S014, run the gel at 140 V for 30 min and then observe in Gel Doc to check the PCR results.

[0107] Among them, the enzyme digestion reaction method includes:

[0108] S011, transformation was performed in commercial DH5α competent cells according to the manufacturer's instructions.

[0109] S012, extract the plasmid according to the instructions of the commercial plasmid DNA mini kit I.

[0110] S013, digest the plasmid according to the instructions of commercial restriction enzymes to linearize the DNA template. Pick a single colony from the culture medium and amplify it in LB medium containing the corresponding resistance. Use Plasmid DNA Mini Kit I to extract the plasmid.

[0111] S014, linearize with appropriate restriction enzyme by incubation at 37°C for 1-2 hours.

[0112] S02, the linearized template of S01 was mixed with an equal volume of chloroform / isoamyl alcohol (24:1 vol / vol), centrifuged at 14,000 g for 1 min at 4 °C, and the upper aqueous phase was carefully separated into a new tube.

[0113] Furthermore, in order to obtain the best separation effect, the DNA template was extracted once more with chloroform / isoamyl alcohol.

[0114] S03, add an equal volume of isopropanol and 1 / 10 volume of 3M NaOAc pH 5.2, immediately snap-freeze in liquid nitrogen, and then centrifuge at 14,000 g for 45 min at 4 °C.

[0115] S04, carefully discard the supernatant, and wash the precipitate twice with 70% (volume / volume) ethanol, dry it naturally until no ethanol remains, and then dissolve the DNA template in 50 μL of DEPC-treated water.

[0116] Measure the DNA concentration for later use.

[0117] S05, IVT with T7 RNA polymerase, includes the following steps:

[0118] S051, mix 30 μg of linearized DNA template in S04, 30 μL of 100 mM NTPs, 100 μL of 10× transcription buffer, 10 μL of 1 M DTT, 60 μL of 7.5 U / μL T7 RNA polymerase, 5 μL of 40 U / μL RNase inhibitor, and 10 μL of 0.1 U / μL pyrophosphatase, and adjust to a total reaction volume of 1 mL with DEPC-treated water in S04.

[0119] S052, incubate the mixture at 37°C for 1-3 hours.

[0120] S053, add DNase I to digest the DNA template at 37°C for 15 min.

[0121] S054, measure the concentration of the IVT product for later use.

[0122] Furthermore, in order to obtain better results, the RNA in vitro transcription and purification method further includes S06 between S05 and S07.

[0123] S06, the MgCl2 concentration was adjusted to 40 mM, and three rounds of thermal cycling were performed (72°C for 1 min, 65°C for 5 min, and 37°C for 10 min each round) to remove the hammerhead ribozyme or HDV ribozyme sequences in the IVT product.

[0124] S07, check the homogeneity of the IVT product by denaturing PAGE and / or check its folding state by native PAGE.

[0125] Wherein, denaturing PAGE comprises the following steps:

[0126] S071, 10 ng of IVT mixture (of S05 or S06) was mixed with 2× TBE urea loading buffer to make a final 2 μL of 1× solution and incubated at 95°C for 3 min.

[0127] S072, load the samples on a 7 M urea 29:1 acrylamide / bisacrylamide gel and run the gel at 300V.

[0128] The concentration of the acrylamide / bisacrylamide solution (29:1) and the electrophoresis time depend on the size of the RNA. For RNAs less than 200 nt, use an 8% gel for 1-2.5 hours; for RNAs between 200 nt and 500 nt, use a 6% gel for 2-3.5 hours; for RNAs greater than 500 nt, use a 5% gel for 3-4.5 hours. Increasing the acrylamide / bisacrylamide ratio to 19:1 results in a smaller pore size, which is more conducive to separating smaller RNAs.

[0129] S073, gel was stained with SYBR Gold and visualized in Gel Doc.

[0130] Wherein, non-denaturing (Native) PAGE comprises the following steps:

[0131] S071, 10 ng of IVT mixture was mixed with 2× native THE loading buffer to make a final 2 μL of 1× solution.

[0132] S072, load the samples on a non-denaturing 29:1 acrylamide / bisacrylamide gel and run the gel at 140 V. The entire process should be performed at 4°C to prevent overheating that may cause RNA denaturation.

[0133] For RNA less than 200 nt, use 8% gel and the electrophoresis time is 0.6-1.5 hours; for RNA between 200 nt and 500 nt, use 6% gel and the electrophoresis time is 0.8-2 hours; for RNA greater than 500 nt, use 5% gel and the electrophoresis time is 1-2.5 hours.

[0134] S073, gel was stained with SYBR Gold and visualized in Gel Doc.

[0135] After evaluation in S07, if the IVT result is ideal, proceed to S07; if the IVT result is unsatisfactory, repeat steps S051-S054 using different conditions (see Table 1) and evaluate again.

[0136] Table 1

[0137]

[0138] S08, buffer exchange or SEC purification to separate different conformations under non-denaturing conditions. These two methods are used to directly obtain RNA samples with different conformations for subsequent S3, RNA sample vitrification, cryo-electron microscopy observation and data collection; or to obtain samples through denaturing gel separation and purification for subsequent S2, RNA folding condition screening.

[0139] Wherein, the method of buffer replacement comprises the following steps:

[0140] S081, add 50 μL of proteinase K (0.6 U / μL) to 1 mL of IVT product (of S05 or S06), incubate at 37°C for 30 minutes to digest proteins, and then centrifuge at 14,000 g for 5 minutes at 4°C.

[0141] S082, transfer the supernatant to a concentration column, replace with 1× IVT buffer, and centrifuge at 12,000 g for 15 min at 4°C. Repeat this process until the OD260 / 280 of the flow-through is less than 20 ng / μL.

[0142] The concentration column is a truncated concentration column, wherein the cutoff is 1 / 3 of the RNA molecular weight. Too small a cutoff will prolong the concentration time, while too large a cutoff will result in the loss of RNA.

[0143] S083, transfer the concentrated sample to a new 1.5 mL tube, measure and label the RNA concentration.

[0144] S084, repeat S07 to check the homogeneity and folding state of the concentrated IVT product.

[0145] Among them, SEC purification separates different conformations under non-denaturing conditions, using an AKTA micro pure instrument and a Superdex increase 3.2 / 300 column, including the following steps:

[0146] S081, wash the column with 1× IVT buffer to balance, set the flow rate to 0.3 mL / min, and the detection wavelength to 260 nm.

[0147] S082, using a peak separation collection method or a fixed volume collection method.

[0148] S083, inject IVT product (S05 or S06).

[0149] S084, use native PAGE to evaluate the RNA conformation composition of samples recovered from different separation peaks. If necessary, use S08 buffer replacement to concentrate samples from different separation peaks to increase the concentration of the samples.

[0150] Wherein, separating and purifying RNA by denaturing gel comprises the following steps:

[0151] S081, add 50 μL of proteinase K (0.6 U / μL) to 1 mL of IVT product (of S05 or S06), incubate at 37°C for 30 min to digest the protein, then centrifuge at 14,000 g for 5 min at 4°C and repeat denaturing PAGE for large-scale denaturing PAGE purification.

[0152] S082, Isolation of RNA:

[0153] Use a handheld UV lamp to cut the target band, then crush the gel into a slurry into a 50 mL tube, add elution buffer, and incubate overnight. Filter the mixture through a 0.22 μm filter. Or

[0154] Perform electrophoresis elution: transfer the gel to the chamber of an electrophoresis elution apparatus and run it at 100 V for 2-12 h in 1× TBE buffer. Transfer the solution containing the target RNA to a new 1.5 mL tube.

[0155] S083, measure concentration.

[0156] Furthermore, RNA in vitro transcription and purification also includes:

[0157] S09, assess the quality of the sample purified by denaturing gel separation by checking the homogeneity of the IVT product by denaturing PAGE and its folding state by native PAGE.

[0158] S01-S04 makes ( Figure 1 (a) The linearized DNA template introduced hammerhead ribozyme and hepatitis delta virus (HDV) ribozyme sequences at the 5' and 3' ends of the target RNA sequence to ensure the uniformity of the IVT product; in the PCR reaction, the 2'-methoxy-modified reverse primer ensured the uniformity of the 3' end.

[0159] S05-S09 enables: IVT products to be rapidly detected by denaturing gel electrophoresis ( Figure 1 bc), and purification of co-transcriptionally folded RNA after protein and DNA template digestion ( Figure 1In the case of RNA fragments that cannot be folded co-transcriptionally (e.g., misfolded conformations), or when excessive byproducts are generated by IVT, mainly due to premature termination, unwanted cleavage events, and internal transcription initiation, large-scale purification using denaturing gel electrophoresis followed by RNA folding ( Figure 1 (f) Low recoveries by conventional gel immersion separations can often be improved by electroelution.

[0160] S2, RNA folding condition screening, including the following steps:

[0161] S10, prepare the reaction system: mix 100 ng RNA, 1 μL of 10× refolding buffer (see Table 1), and DEPC-treated water to make the total reaction volume of each screening condition 9 μL.

[0162] Screening of RNA refolding conditions by PAGE.

[0163] S11, prepare the stock solution corresponding to the MgCl2 concentration (see Table 1), and then add 1 μL of MgCl2 solution to each condition according to the cooling strategy (see Table 2).

[0164] Table 2 Different cooling strategies

[0165]

[0166] S12, Repeat the folding state by native PAGE to evaluate RNA refolding conditions.

[0167] S13. Adjust the volume according to S11 to obtain 20 μL of folded RNA per screening condition, then centrifuge at 15,000 × g for 15 min at 4 °C.

[0168] To obtain ideal results from nanoDLS, the final RNA concentration should be between 100 and 500 ng / μL and meet the nanoDLS parameters - scattering intensity above 106 cts / s and SNR above 120 (for Prometheus Panta instrument).

[0169] S14, load 10 μL of the folded RNA solution (product of S11) into the capillary. Two capillaries were used for each condition.

[0170] S15, place the capillary tube on a Prometheus Panta tray (NanoTemper Technologies).

[0171] S16, running Prometheus Panta size analysis in high sensitivity mode.

[0172] S17, data were analyzed using Prometheus Panta analysis software—observation of differences in intensity distribution curves, rH, and PDI.

[0173] RNA folding screening also includes:

[0174] S18, for stacked intensity curves, raw size analysis data were exported using Panta Control software and analyzed and plotted using Origin software.

[0175] In S12, if the conditions tested are not ideal and more variables need to be screened extensively, repeat S10-S12 or perform screening at a higher throughput using nanoDLS according to S13-S18.

[0176] S13-S18 make:

[0177] Folding condition screening involves preliminary screening of different heating / cooling strategies and buffer compositions, followed by expanded screening of pH ranges, ion concentrations, and additives, with the goal of finding the best folding conditions that allow RNA to fold into uniform conformations, which are detected by non-denaturing PAGE (Figures 1c and 1g, Tables 1 and 2). Although a band does not necessarily contain a single conformation, multiple bands usually correspond to multiple conformations. Blurred bands indicate that the RNA has a loose structure or multiple dynamic conformations, and its conformational exchange rate is lower than the electrophoretic mobility. Most RNA denatures at 90°C, while GC-rich RNA may require higher denaturation and folding temperatures. Cooling strategies for different time periods, addition of Mg 2+ Different time points of Mg ions and extended incubation times in different folding buffers can lead to different states of the same RNA. Increasing RNA concentration may lead to RNA aggregation. In order to obtain enough particles in the cryo-EM field of view, a higher RNA concentration is usually required, and aggregation can be avoided by folding at low RNA concentrations and then concentrating and centrifuging. Although Mg 2+ It is essential for the correct folding of RNA, but high concentrations of Mg 2+ May cause aggregation of RNA, thus significantly reducing the number of monodisperse particles - reducing Mg 2+concentration may alleviate this problem. Aggregation of RNA may also be beneficial (especially for smaller RNAs) because the increase in molecular weight often facilitates particle extraction and fitting of 2D classifications. NanoDLS measures particle size and its polydispersity by monitoring the fluctuations in the intensity of scattered light as particles undergo Brownian motion in solution. If existing IVT and RNA folding conditions do not yield uniform conformations, nanoDLS can be used to systematically examine more variables, such as different ions and additives, at higher throughput ( Figure 1 h). The optimized conditions can be used for molecular sieving SEC to further separate different conformations and / or RNA aggregation states, followed by subsequent vitrification and cryo-EM observations.

[0178] Example 3 Solving RNA heterogeneity

[0179] Based on the RNA in vitro transcription and purification and RNA folding of Example 2, this Example 2 provides a method for solving RNA heterogeneity.

[0180] A method for resolving RNA heterogeneity, comprising the following steps:

[0181] S3, RNA sample vitrification, cryo-EM observation and data collection, including the following steps:

[0182] S19, repeat S10-S11 in Example 2 to prepare the required amount of RNA samples under optimal conditions for vitrification and cryo-electron microscopy (cryo-EM) analysis.

[0183] S20, set the temperature and relative humidity of the Mark IV Vitrobot to 4 °C and 100%, assemble the filter paper in the chamber and cryogen container for vitrification.

[0184] S21, select the appropriate type of grid (see Table 3 and Table 4) and perform glow discharge. Copper grids are mainly used in the screening stage. Support membranes can be used when the final concentration is less than 1 μM (or 0.1 μg / μL) and the RNA is larger than 100 kDa (300-nt).

[0185] Table 3 Different types of grids

[0186]

[0187] Table 4 Different types of support films

[0188] Graphene Graphene oxide Ultrathin carbon 2D crystals of streptavidin and hydrophobin HFBI Hole ceramic

[0189] Grids without a support film are typically glow discharged for 40-120 seconds, while grids with a support film are typically glow discharged for a shorter time, usually less than 10 seconds, depending on the strength of the film. Glow discharge is not recommended for certain extremely fragile films.

[0190] S22, load the grid onto the Vitrobot, add 2-3 μL of sample onto the grid, consider adding a detergent at 10% critical micelle concentration to help particles enter the wells (Table 5). Use filter paper to absorb excess sample and inject a cryogen (e.g., liquid ethane). Transfer the grid to a grid box pre-cooled with liquid nitrogen. Clip the automated grid for future use.

[0191] The grid without a support film usually absorbs water for 2-3 seconds, while the grid with a support film usually waits 30-60 seconds and then absorbs water for 1-2 seconds to allow particles to accumulate on the membrane.

[0192] Table 5 Common detergent types that help RNA particles enter the well

[0193]

[0194] S23, load the automated grid into a pre-cooled box, transfer the box into a NanoCab filled with liquid nitrogen, and dock the box onto the microscope. Check that the microscope is ready for imaging.

[0195] S24, open the EPU and select Atlas mode to get an overview of the ice distribution on the grid; move the sample stage to the area of ​​interest and acquire images in GridSquare mode; find the concentric height, move the sample stage to the hole with the appropriate ice layer, and enter Hole / EucentricHeight mode; collect cryo-EM images in Data Acquisition mode.

[0196] S25, if the grid is not ideal and is not suitable for data collection, remove the grid and repeat S24 to screen the next grid.

[0197] S26. After completing the screening of all automatic grids, if the ice thickness is not ideal, repeat S20-S25 to screen the grids, glow discharge time and Vitrobot parameters until the optimal ice thickness is obtained; if the RNA sample state is not suitable for data collection (e.g., low concentration, aggregation, denaturation), repeat S10-S19 to screen variables in IVT and RNA folding until the best micrographs suitable for data collection are observed.

[0198] S27, Activate the "Ice Filter" in EPU to select wells based on ice thickness determined by image intensity. Critical Steps Ice thickness can also be quantitatively measured in real time using methods such as MeasureIce integrated in EPU.

[0199] S28, Set up automatic data collection according to the EPU protocol and check the microscope and camera settings to ensure that data collection proceeds smoothly.

[0200] Among them, multiple exposures can be set to increase the throughput of data collection; the sample stage can be tilted to collect data on a grid with a preferred orientation; the direct electron detection camera is usually set to counting mode for data collection. For K2 and K3 cameras (Gatan), super-resolution mode can be selected to increase the sampling frequency of micrographs, but the size of the micrographs will increase significantly compared to the photos in counting mode; automatic data transfer to a data processing cluster or an external hard disk can be set to avoid overfilling the local storage space.

[0201] S19-S28 state that the optimal vitrified grid should have monodisperse particles randomly oriented in as thin an ice layer as possible, ideally with an ice layer thickness that only covers the largest dimension of the particles. Too thin an ice layer may result in particle rejection, particles with preferred orientation, or ice breakage during imaging, while too thick an ice layer results in multilayering of particles and increased inelastic scattering, which reduces the signal-to-noise ratio (SNR) and reconstruction resolution (S19-S22). The main factors affecting the quality of vitrified samples include ice thickness, particle distribution, sample denaturation, and preferred orientation of particles. While the first two issues can be observed directly on cryo-EM micrographs, the latter two issues usually require the collection of small test data sets and data processing until 2D classification or later steps are performed. All of the above issues can be circumvented by adding detergents to the sample and by choosing different types of grids and support films (Tables 3 to 5), glow discharge / plasma cleaning equipment, and cryostat. Detergents added at concentrations at and below the critical micelle concentration (CMC) have been shown to affect ice thickness and particle distribution. In addition, they can improve the preferred orientation and sample denaturation at the air water interface (AWI), which attracts and damages macromolecules during the freezing process. Various types of grids with different grids and film materials other than conventional copper grids and carbon films are recommended to minimize particle motion, ice thickness, surface tension, and preferred orientation. Adding support films can help particle adsorption on the membrane, improving ice thickness and particle distribution. At the same time, these membranes can block AWI and adsorb particles, thereby improving preferred orientation and sample denaturation. However, the most commonly used amorphous carbon films produce significant background noise and may not be suitable for small RNA samples and near-atomic resolution cryo-EM reconstructions. Thinner support films, such as graphene, graphene oxide, reduced graphene oxide, two-dimensional crystals of streptavidin, and hydrophobic protein HFBI, can be deposited on most types of grids as an alternative. A list of detergents, grids, and support films is summarized in Tables 3 to 5. Grids with or without support films are usually hydrophobic and require glow discharge or plasma cleaning. Devices that generate residual air plasma are widely used (e.g., PELCO easiGlow and Tergeo-EM), primarily because of their low cost and ease of operation, to remove surface contaminants and increase hydrophilicity, resulting in a more uniform distribution of aqueous samples. Alternative devices can utilize low-pressure gas mixtures in specific combinations, such as argon-oxygen (e.g., Solarus), or vaporized chemicals such as amylamine (GloQube). Special care should be taken with fragile support membranes like graphene, which may be damaged during the plasma cleaning process. Vitrification preserves purified macromolecules in a near-native state by rapidly freezing a thin layer of solution to form amorphous ice.Several manual and semi-automated cryoplunge devices are widely used, which remove excess water sample through filter paper and immerse the sample in a cryogen (such as liquid helium). Common ones include the Vitrobot Mark IV (exemplified in this method), Leica EM GP2, and Gatan Cryoplunge 3. Recently, several highly automated devices have been developed to minimize manual operations and avoid the use of filter paper blotting, using methods such as inkjet dispensing, spraying, needle printing, microcapillary writing, or tube aspiration. The vitrified grid is loaded into the microscope for cryo-EM observation (S23-S26). Appropriate ice thickness is crucial for obtaining high-resolution cryo-EM data, however, all of the above vitrification methods will produce different ice thicknesses even within the same well of the grid. Therefore, it is crucial to screen the area with the optimal ice thickness before semi-automated data acquisition (S27-S28;. Figure 2 (a) Commonly used ice thickness measurement methods include tomography, additional imaging techniques combined with machine / deep learning, and real-time methods using energy filters or objective apertures, some of which can be integrated in the most commonly used automated data acquisition schemes such as EPU, SerialEM, and Leginon.

[0202] S3, Data processing to address RNA heterogeneity, includes the following steps:

[0203] S29, import all cryo-EM images and perform "MotionCorrection" using MotionCor2, set "Number of patches X,Y" to "5,5", and set "Other MotionCor2 parameters" to "-Iter 10". The dose per frame is provided accordingly, and the rest of the parameters remain default.

[0204] For relatively small RNA or sparse particle distribution, "Number of patches X,Y" can be set to "3,3". MotionCor2 is also integrated in cryoSPARC, and other movie alignment methods in Relion, cryoSPARC, Warp, EMAN2, and cisTEM can also be used to evaluate motion in the whole frame or localized regions.

[0205] S30, perform "Contrast transfer function (CTF) estimation" using CTFFIND4 with default settings, but set "Amount of astigmatism" to Contrast ring out of The micrographs were visually inspected and selected.

[0206] For tilted data, the latest CTFFIND5 (integrated in cisTEM) and other methods such as Gctf (integrated in Relion) or patch CTF (integrated in cryoSPARC) can be used.

[0207] S31, using Blob picker in cryoSPARC, reference-based template matching in Relion, and reference-based CNN particle picking in EMAN2.

[0208] Using the Blob picker in cryoSPARC includes the following steps:

[0209] S311, start a new "Blob Picker" job and import the micrographs from S30.

[0210] S312, check "Show advanced" and set the expected minimum and maximum particle diameters. For Tetrahymena group I introns, this value is set to 50 and

[0211] S313, open "Circular blob" and run the job, keeping the rest of the parameters as default settings.

[0212] S314, after completing the "Blob Picker" job, start a new "Inspect Particle Picks" task. Import the photos and particles from the "Blob Picker", keep the parameters as default, and adjust the "Power Histogram" to visually inspect the particle picking results for each micrograph.

[0213] S315, when the selected particles are as many as possible and uniform, click "Done Picking".

[0214] Picking too few particles will result in the loss of valid particles, which may affect the reconstruction resolution, while picking too many particles will increase the data processing time.

[0215] Among them, using reference-based template matching in Relion includes the following steps:

[0216] S311, import the micrographs from S30 and set the corresponding “Pixel size in micrographs”.

[0217] S312, import the results of the two-dimensional classification average, and set the "Lowpass filter references" to Set the Pixel size in references as required.

[0218] Among them, using reference-based CNN particle picking in EMAN2 includes the following steps:

[0219] S311, navigate to "Particles->Interactive Particle Picking" in the GUI and import the micrograph from S30. Interactively enter the "boxsize" and "ptclsize" values ​​determined from the micrograph. For the Tetrahymena group I intron, these values ​​are set to 256 and 200. Fill in the remaining parameters.

[0220] S312, manually select 15-20 reference images as good and background references.

[0221] The reference should be selected from different micrographs and have different defocus values.

[0222] S313, train the CNN model using the “Threshold1” value set to 2.

[0223] S314, click "Autobox" on a single micrograph to check the particle picking results.

[0224] Adjust the value of Threshold1 and click Autobox to interactively view the results.

[0225] S315, when the selected particles are as many as possible and uniform, click "Autobox All".

[0226] S32, perform "grain extraction" using the imported micrographs in S30 and the "input coordinates" generated in S31. Set an appropriate "grain box size" (256 pixels for Tetrahymena group I introns) and scale it to 64 pixels by a factor of 4 to speed up the process of two-dimensional average classification (2D classification) and three-dimensional average classification (3D classification) of grains.

[0227] S33, perform "2D classification" on the scaled particles. Set "Number of classes" to 200, "Use EMalgorithm" to "Yes", "Number of EM iterations" to 50, "Use VDAM algorithm" to "No", and "Mask diameter" to and set Limit resolution E-step to Keep the rest of the parameters as default.

[0228] S34, visually inspect the classification results in “Subset selection” and discard those unwanted RNA particles and classes without obvious RNA features.

[0229] S35, repeat S33-S34 for 2-3 times until most of the unwanted particles are discarded. If no RNA feature is observed, the data processing is terminated and S10-S34 is repeated until the 2D classification results reveal the RNA feature.

[0230] S36, import particles from S35 and generate a "3D initial model", set the "Number of classes" to 6, and keep the other parameters as default or consistent with the previous steps.

[0231] Equivalent to the "Ab-Initio Reconstruction" job in cryoSPARC, faster reconstruction can be performed using partial particles.

[0232] S37, perform "3D classification" by importing all particles in S35 and a "reference image" in S36. Set the "Number of classes" to 6 and the "Limit resolution E-step" to Keep the rest of the parameters as default.

[0233] This is equivalent to "3D Classification" in cryoSPARC. Multiple initial models can also be provided in this step, similar to "Heterogeneity Refinement" in cryoSPARC. Homology models and prediction models can be converted to cryo-EM density maps and applied The resolution of the image is filtered and used as the initial model, which can be set in "Initial low-pass filter".

[0234] S38, repeat S34 to select particles from the desired category, and then repeat S32 to extract the particles without scaling.

[0235] S39, performing preliminary "3D automatic refinement" by importing the selected particles in S38 and a "reference image" in S37.

[0236] Import the “Reference mask” from this step and set “Use solvent-flattened FSCs” to “Yes” to improve the reconstruction of low molecular weight and low signal-to-noise particles.

[0237] S40, perform spectrum sharpening in "Post-Processing" and import the "half unfiltered spectrum" and "solvent mask" in S39.

[0238] This is equivalent to "homogeneous refinement" in cryoSPARC, while "local refinement" and "inhomogeneous refinement" can also be used for comparison.

[0239] S41, cryo-EM map visualized in Chimera.

[0240] S42, repeat S37-S41 to obtain a finer resolution of the flexible region density.

[0241] S43, repeat S39-S41 for the selected class of particles and provide the corresponding "reference mask" to obtain the final structure that better resolves RNA heterogeneity.

[0242] New algorithms recently developed in RELION, “3D multi-body” and cryoSPARC, “3DFlex”, allow the capture of continuous motion of multiple flexible regions. These regions may need to be separated independently by the user and each region may have a lower molecular weight limit to ensure optimal performance of the algorithm with sufficient signal-to-noise ratio (SNR).

[0243] S44, import "CTF refinement", "grains" and "post-process STAR files" from S43, first "Estimate magnification".

[0244] S45, import "particles" from S44 into "CTF refinement", and perform "CTF parameter fitting". Set "Fit defocus" and "Fit B factor" to "per particle", "Fit astigmatism" to "per micrograph", and "Esti mate beamtilt" and "Also estimate trefoil" to "yes".

[0245] S46, import "micrographs" from S29, "particles" from S44, and "post-processed STAR files" from S43 into "Bayesian refinement". Set "Train optimal parameters" to "Yes", set "Perform particle refinement" to "No", and keep other parameters as default.

[0246] S47, in "Bayesian Refinement", import the "Optimization Parameters File" from S46, set "Train Best Parameters" to "No", set "Perform Granular Refinement" to "Yes", and keep other parameters as default.

[0247] In cryoSPARC, particles from S43 can undergo the new tasks of “global CTF refinement” and “reference-based motion correction” for similar particle processing.

[0248] S48, repeat S39-S41 to obtain the final structure from the refined particles.

[0249] S49, analyze the "Local Resolution" of the final map by providing the required input files and the "User-provided B-factor" from S48, and the corresponding "Calibrated Pixel Size". Set "Use Relion" to "Yes" and import the "Detector MTF (STAR ​​File) detector MTF" from the corresponding detector manufacturer.

[0250] S50, Local resolution density map of S49 colored for presentation in Chimera.

[0251] Among them, open "relion_locres_filtered.mrc" in Chimera; select "Tools->Volume Data->Surface Color"; select "Surface Color" mode as "By Volume Data Value" and import the volume file "relion_locres.mrc"; select "Options" to modify the parameters. Select "Set" to set the full range of surface values ​​with corresponding resolution. Select "Invert" to adjust the color-resolution relationship; select "Color" to generate a color map of local resolution.

[0252] S29-S50 allow for: Instant assessment of cryo-EM SPA data quality with minimal need for user / manual input, maximizing instrument efficiency and productivity. This is typically followed by extensive data processing in order to resolve functionally relevant flexible regions, dynamic conformational changes, and sample composition heterogeneity, and to improve resolution. In addition to standard data processing pipelines (S29-S41), the integration of deep learning algorithms helps improve the picking of low-contrast particles, as well as classifying continuous conformations, allowing the inference of functionally relevant motion pathways. These features are very useful for resolving RNA dynamics and heterogeneity. The data processing protocol is illustrated using examples using the commonly used software suites EMAN2, Relion, and cryoSPARC (S42-S50).

[0253] Example 4 RNA modeling and manual revision and new modeling based on RNA structure prediction

[0254] Based on the method for solving RNA heterogeneity in Example 3, this example provides a method for RNA modeling and manual revision and new modeling based on RNA structure prediction.

[0255] A method for RNA modeling and manual revision and new modeling based on RNA structure prediction, comprising the following steps:

[0256] S5, RNA modeling into cryo-EM density, includes the following steps:

[0257] S51, search RNA sequences on the Rfam sequence search page https: / / rfam.org / search#tabview=tab1 to find homology models. First, input RNA sequences for search.

[0258] S52, navigate to the discovered Rfam family and click on the "Structures" tab on the left to view the results. Download the model via the link provided. If no homology model is found, proceed to the manual revision and new model stage S60.

[0259] S53, open the homology model and cryo-EM density map in Chimera and manually fit the model to the density map so that most parts of the model lie inside the density.

[0260] Make sure the model and density plot are fully overlapped in three dimensions by turning on "Tools->View Controls->Side View".

[0261] S54, open "Favorites->Command Line", and use the following command to perform rigid body fitting:

[0262] fitmap#1#0

[0263] Where #1 represents the model and #0 represents the density map.

[0264] S55, repeat S53-S54 until most parts of the model fit the density map.

[0265] S56, manually select the regions in each model that are still outside the density, navigate to “Tools->Volume Data->Fit in Map”, and fit the “Selected Atoms” in the specified “Map”.

[0266] To ensure that unselected regions of the model that have been fit to the density are not moved, open "Options" in the "Fit in Map" module and deselect the "Move entire molecule" option.

[0267] S57, navigate to "Tools->Move->Move Mouse Model->Move Selection", manually rotate and move the selected area into the density, and then repeat S56.

[0268] S58, navigate to "File->Save PDB", check "Save relative to model", and select the specified density plot to save the fitted model.

[0269] It is important to ensure that the model still fits the density map when opened in other software.

[0270] S59, open the model and density map from S58 in Coot and navigate to "Compute->Model / Fitting / Refinement->Simple mutations…" to generate the correct RNA sequence. If the corrected model is complete and fits the cryo-EM density well, you can directly proceed to the model refinement and validation phase starting from S72.

[0271] S6, Manual revision and new modeling based on RNA structure prediction, including the following steps:

[0272] S60, open the model of S59 in Chimera. If the model is incomplete or segmented, contains improper RNA folding or does not fit well with cryo-EM density, manually select and remove all improper regions by navigating to "Operations->Atoms / Bonds->Delete".

[0273] S61, predict missing regions using software or online servers such as SimRNA, FARFAR, or RNAcomposer. Provide sequence and secondary structure as input.

[0274] If the predicted model contains a broken sequence, such as an internal stem, it is recommended to insert a terminal four-leaf loop sequence (such as GAAA) into it for structure prediction and then remove it from the generated model as described in S60.

[0275] S62, and S54–57 were repeated to fit all missing regions into the cryo-EM density based on RNA sequence and secondary structure.

[0276] In extreme cases where most or all regions are poorly modeled, manual new modeling will be very challenging and time consuming. A practical starting point is to identify multi-way junctions in the secondary structure and correlate with the density map. Adjacent stems can be predicted and fitted into the density individually. The accuracy of the secondary structure has a significant impact on new modeling time.

[0277] S7, troubleshooting, includes the following steps:

[0278] S63, Open all saved models and density maps in Coot. Change the residue numbers according to the sequence in "Edit->Renumber residues…".

[0279] S64, merge all fragments in "Edit->Merge molecules..."

[0280] S65, change all fragments to the same chain ID in "Edit->Change Chain ID..."

[0281] S66, navigate to “Compute->Model / Fitting / Refinement->Add terminal residues…” to add the missing residues to ensure that the complete sequence is modeled.

[0282] S67, check each residue in the full RNA model one by one to ensure that it fits correctly into the density by navigating to "Go to atom...->Next residue". In "Compute->Model / Fit / Refinement->Real space refinement area", residues outside the density can be activated and then manually adjusted to reposition inside the density.

[0283] Check the N-glycosidic bonds of each nucleotide to ensure that they are mostly in the more stable trans conformation. Occasionally, less stable cis conformations are observed and need to be verified by high-resolution density.

[0284] S68, when the resolution is close to or better than When , metal ions are resolved and can add to the extra density, especially near phosphate oxygens. Larger extra density may be attributed to small molecules such as spermidine that are added during IVT or RNA refolding.

[0285] S69, after checking, save the entire model for subsequent refinement.

[0286] S8, automatic and semi-automatic modeling, includes the following steps:

[0287] S70, if no homology model is available, a model can be automatically generated. Several automatic and semi-automatic modeling tools, such as phe nix.map_to_model, ModelAngelo, CryoREAD, EMRNA, and auto-DRRAFTER, use RNA sequence and secondary structure as input to generate atomic coordinates from cryo-electron microscopy density maps at sub-nanometer to near-atomic resolution.

[0288] Among them, Phenix.map_to_model includes the following steps:

[0289] S701, navigate to "Model Building->Graph to Model" and open the "Graph to Model" interface in Phenix.

[0290] S702, import the map and sequence, and fill in the "high resolution restraints" from the cryo-electron microscopy map.

[0291] S703, if a starting model is provided, the accuracy and speed of model building will be improved.

[0292] S704, click “Run” to execute the task, and the output file “map_to_model.pdb” will be generated.

[0293] Among them, ModelAngelo includes the following steps:

[0294] S701, navigate to "Relion 5.0->ModelAngelo Build".

[0295] S702, import the "B-factor sharpening map", "protein FASTA sequence" and "RNA FASTA sequence" from S40 as input.

[0296] Protein sequences are required as a required input, even if no protein is present in the structure.

[0297] S703, RNA secondary structure can be provided as an additional input file in dot-bracket format.

[0298] S704, set "Execute HMMer search" to "No".

[0299] S705, click “Run” to execute the task with GPU acceleration.

[0300] Among them, CryoREAD includes the following steps:

[0301] S701, install CryoREAD from GitHub.

[0302] S702, executing commands to detect possible positions of phosphates, sugars, and bases based on a deep neural network.

[0303] S703, open the prediction results in Chimera to check the positions of phosphates, sugars, and bases.

[0304] S704, executing a command to build a model without an RNA sequence, or executing a command to build a model with an RNA sequence.

[0305] S705, repeat S703 to visualize the model.

[0306] Among them, EMRNA includes the following steps:

[0307] S701, install EMRNA from the designated website.

[0308] S702, prepare RNA secondary structure, expressed in dot-bracket format.

[0309] S703, executing commands to build a model.

[0310] phenix.map_to_model in Phenix and ModelAngelo in Relion provide graphical user interfaces, while running CryoREAD and EMRNA requires command lines. Direct use of automatically generated RNA models without manual revision remains challenging, especially for cryo-EM density maps with flexible regions and low local resolution. For additional information and advanced options of the automated modeling software, see their websites.

[0311] S71, Check the RNA model generated from S70 in Chimera. If the model is incomplete or fragmented, contains improper RNA folding or does not fit well to the cryo-EM density, repeat S60-S69.

[0312] S8, model refinement and validation, includes the following steps:

[0313] S72, open the Phenix GUI and select phenix.real_space_refine. Import the density map and model from S69 or S71 and enter the resolution of the cryo-EM map.

[0314] S73, in the Refinement Settings tab, check "Deformation", uncheck "nqh_flips" and "Ramachandran Constraint", and keep other parameters as default. Set "Root Mean Square Deviation of Target Key" to 0.002 and "Root Mean Square Deviation of Target Angle" to 0.2.

[0315] S74, click "Select Atoms->Find Secondary Structure" to generate a secondary structure text file. Open the file in a text editor and repeat S67 to manually incorporate the missing pairing and stacking information into the file.

[0316] The accuracy and completeness of the text files are critical for model refinement, especially in flexible regions with low-resolution density. Information is required for every base pair.

[0317] S75, load the file of S71 in the real space refinement window.

[0318] S76, click “Run” to start the refinement process.

[0319] S77, Review the Validation Summary page and display the model issues by navigating to "Open in Coot".

[0320] S78, activated problematic residues, and manually adjusted until all parameters were within standard ranges.

[0321] S79, repeat S72-S78 until all outlier residues are fixed, the correlation score between the model and the graph is as high as possible, and the conflict score is as low as possible.

[0322] Generally, CCmask values ​​over 0.75 and Clash Score below 5 are acceptable. RNA models can be further refined and regularized by other software.

[0323] S80, navigate to the relevant website for verification. Upload the coordinate file and follow the instructions to add hydrogen atoms. Execute the "Analyze All-Atom Contacts and Geometry" function. Repeat S78 until there are no major issues in the model.

[0324] S81, navigate to the relevant website, upload the cryo-EM images and refined models, and follow the verification instructions. Repeat S77-S78 until there are no major issues in the verification report.

[0325] From S51-S81, we can see that:

[0326] Building atomic models of cryo-EM density has traditionally been time-consuming and requires expertise and extensive experience. The model is usually built with confidence when the resolution is reduced to near When the model is smaller or worse, the confidence level of the model decreases and the errors in the model increase. Despite recent major advances in accurate structure prediction of millions of protein complexes and highly automated modeling in high-resolution cryo-EM reconstructions, atomic modeling of cryo-EM density at intermediate resolution remains challenging.

[0327] Although some RNA cryo-EM structures have achieved overall resolution and better, but the flexible regions in these structures are usually at lower resolution. In addition, most RNAs are still at a medium resolution range. Although several automated methods have been developed to model RNA into cryo-EM density, they usually still require a lot of manual intervention, RNA structure prediction and subsequent revision. We propose a general method that starts with docking of homology models, followed by manual revision and de novo modeling, which can be assisted by structure prediction and automated modeling algorithms (steps 70-71). The final model is revised and verified before being deposited in a public database.

[0328] Example 5

[0329] The methods formed in Examples 2 to 4 are compared with the existing methods, and the results are as follows:

[0330] (1) Alternative structural biology methods

[0331] Cryo-EM single particle analysis (cryo-EM SPA) requires minimal sample and is rapidly frozen in vitreous ice to preserve the native conformation. Evolving algorithms allow for better resolution of RNA dynamics and conformational changes. Cryo-EM SPA complements alternative structural biology methods such as X-ray crystallography and NMR, which provide atomic-scale RNA coordinates, while small-angle X-ray scattering (SAXS) and atomic force microscopy (AFM) provide low-resolution contour and shape information.

[0332] (2) X-ray crystallography

[0333] X-ray crystallography requires large sample sizes and successful growth of RNA crystals to high resolution, which is clearly challenging for many RNAs. Therefore, sequence engineering based on extensive sequence alignments, functional studies, and biophysical / biochemical characterization to minimize perturbations to the tertiary structure still plays an important role in optimizing crystallization conditions, crystal packing, and diffraction quality. The design of RNA constructs includes phylogenetic analysis to achieve truncation of functional RNA fragments, introduction of GNRA tetraloops, U1A and / or L7Ae binding motifs, directed mutagenesis, and / or systematic mutational evolution. In contrast, cryo-EM allows for direct visualization and determination of the structure of wild-type, full-length RNA.

[0334] (3)NMR

[0335] NMR is able to resolve the structure of RNA and its dynamics by measuring the chemical shift of nuclear spins under biologically relevant conditions. However, as RNA increases in size, spectral signals overlap and expand, and signal intensity decreases, which becomes a major limiting factor in RNA structure determination. Therefore, the average size of RNA structures determined by NMR is 30-nt. Nucleotide-specific labeling strategies developed for the study of larger RNAs have been successfully applied to several RNAs up to 155-nt. However, the required sample amount is also large, and determining the overall folding of large RNAs is often challenging.

[0336] (4) SAXS and AFM

[0337] Small angle scattering determines the average shape of RNA in solution through scattering events generated by X-ray or neutron beams (SANS). The resolution of the electron density map obtained by SAXS and the atomic distribution obtained by SANS ranges from 1 to 25 nanometers. AFM uses a cantilever with an ultrasonic probe to scan a stationary sample, which reacts with the sample in touch or contact mode to move a piezoelectric element. The changes caused by the movement of the laser reflection can be detected by a photodetector, and the detected signal is then converted into surface information. The accuracy can be improved by optimizing the sample fixation and using sharper probes. Both methods can evaluate RNA samples at nanometer resolution and provide information on the overall shape of the RNA, but cannot obtain high-resolution details compared to other structural biology methods.

[0338] (5) In situ structure determination

[0339] In situ cryo-EM SPA and cryo-electron tomography (cryo-ET) combined with focused ion beam (FIB) milling have allowed near-atomic-scale structure determination of biomolecular complexes in the native cellular environment. However, in situ structure determination is currently limited to large biomolecules and complexes (>1 MDa) that are highly abundant within cells. Furthermore, the availability of automated FIB milling instruments remains limited.

[0340] (6) The method of the present invention has

[0341] The RNA preparation section generally applies to all in vitro structure determinations. In vitro transcription (IVT) and RNA refolding conditions often need to be screened and optimized individually to obtain the most thermodynamically stable RNA conformation.

[0342] Non-denaturing gel electrophoresis is widely available in most laboratories and is used to evaluate RNA under different conditions. Acrylamide and bis-acrylamide polymerize to form a mesh-like gel matrix that can separate RNAs that adopt different conformations or form complexes with different substrates, resulting in different electrophoretic mobilities. The migration of RNA molecules is mainly determined by the dynamics of RNA conformational exchange and therefore cannot be easily predicted by physical properties alone.

[0343] To increase the versatility of the current method, we incorporated nanoDLS into the condition screening step, which provides information on the polydispersity and hydrodynamic radius of folded RNA, and has the option to analyze different samples / conditions in parallel to increase throughput. Other methods that provide additional information about RNA can also be used to enhance the versatility of the protocol, including analytical ultracentrifugation (AUC) and capillary electrophoresis (CE). AUC measures the sedimentation rate of RNA samples in solution to provide information on particle size, shape, and diffusion; CE can separate different RNA conformations and lengths based on the migration rate of RNA through a capillary under an electric field with high sensitivity.

[0344] Example 6

[0345] The method formed in Examples 2-4 is applied to the following:

[0346] Suitable for RNAs of 70-nt or greater that form three-dimensional structures.

[0347] The applicability to large RNAs and ribonucleoproteins (RNPs) makes this method potentially applicable to functional long noncoding RNAs (lncRNAs), especially those structures that have been studied by other methods. For example, small-angle X-ray scattering (SAXS) revealed the overall shape of Bvht before and after binding to its protein partner CNBP, which regulates the expression of key genes related to cardiovascular development; HOTAIR (HOX transcription antisense RNA, which has an oncogenic role in a variety of cancer cells) and MEG3 (human maternally expressed gene 3, downregulated in most cancer cells) formed flexible elongated shapes by atomic force microscopy (AFM); the major small RNA (miR) cluster miR-17 / 92, which contains six cancer-related mi RNAs, also known as oncomiR-163, has been found to form a compact spherical structure under negative staining electron microscopy; 7SK noncoding RNA forms a core RNP with MePCE and Larp7 and can regulate RNAPII transcription, and also plays an important role in pathogenic diseases such as HIV infection. Cryo-EM revealed conformational changes in the truncated 7SK core RNP between “linear” and “circular” forms. Additional cryo-EM analysis of the full-length or isolated functional domains of these lncRNAs may provide mechanistic insights into their functions.

[0348] Many noncoding RNAs perform their functions in the presence of other binding partners (mainly proteins, sometimes RNA or small molecules). Studying RNA monomer structures helps to elucidate the conformational changes that occur during binding and complex formation. For example, the viral tRNA-like structure (TLS) of bromomovirus (BMV) was found to undergo significant conformational changes after binding to the host tRNA synthetase; the group I intron of Cryptosporidium may undergo large conformational changes during substrate binding and catalysis; the SAM-IV riboswitch preforms the binding pocket of the S-adenosylmethionine (SAM) binding site and remains unchanged after binding to SAM.

[0349] RNA as a drug target is particularly important in the pharmaceutical industry because the number of RNAs transcribed from the human genome outnumbers the number of proteins translated by a factor of 50. A recent study identified small molecules that can modulate dynamic conformational changes in the A-containing repeat (RepA) in the RepA subdomain of Xist, a lncRNA that induces stable silencing of the entire chromosome by homotypic accumulation on the inactive X chromosome, thereby achieving dosage compensation between sexes. In addition, Risdiplam, recently approved by the U.S. Food and Drug Administration (FDA), treats spinal muscular atrophy (SMA) by targeting SMN2 pre-mRNA and correcting alternative splicing. This approach can help elucidate the structural basis of RNA-targeted small molecules by cryo-EM, complementing X-ray crystallography and NMR and facilitating structure-based RNA-targeted drug discovery.

[0350] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0351] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present application and form different embodiments. For example, in the above claims, any one of the claimed embodiments may be used in any combination. The information disclosed in this background technology section is intended only to deepen the understanding of the overall background technology of the present application and should not be regarded as an admission or in any form of implication that the information constitutes prior art known to those skilled in the art.

Claims

1. A method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA, characterized in that The following steps are involved: S3, the purified folded RNA is then cryo-vitrified and evaluated under cryo-EM to obtain a grid with optimal ice thickness and uniform particle size and distribution; S4, a small amount of data is collected on the grid. Data processing is paused after two-dimensional (2D) classification is completed for preliminary evaluation. If there are obvious RNA features in the preliminary evaluation, large-scale cryo-EM data collection and subsequent data processing are performed to resolve RNA heterogeneity. If there are no obvious RNA features, different RNA preparation and / or vitrification conditions need to be tried to obtain optimized RNA samples and small amounts of data are collected again until there are obvious RNA features. Large-scale cryo-EM data collection is then performed before data processing is performed to resolve RNA heterogeneity. S5, Most RNAs undergo conformational changes when performing their functions. The standard cryo-EM data processing pipeline combined with the 3D classification algorithm can more effectively separate conformational changes.

2. A method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA, characterized in that The following steps are involved: S3, RNA sample vitrification, cryo-EM observation, and data collection; S4, Data processing to address RNA heterogeneity.

3. The method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA according to claim 2, characterized in that: S3 includes the following steps: S19, prepare the required amount of RNA samples under optimal conditions for vitrification and cryo-electron microscopy analysis; S20, the temperature and relative humidity of the Mark IV Vitrobot were set to 4 °C and 100%, and filter papers were assembled in the chamber and cryogen container for vitrification; S21, select the appropriate type of grid and perform glow discharge, wherein the copper grid is mainly used in the screening stage, and the support membrane is used when the final concentration is lower than 1 μM and the RNA is greater than 100 kDa; S22, load the grid onto the Vitrobot, add 2-3 μL of sample onto the grid, use filter paper to absorb the excess sample and inject cryogen, and transfer the grid to the grid box pre-cooled with liquid nitrogen; S23, load the automated grid into a pre-cooled box, transfer the box into a NanoCab filled with liquid nitrogen, and then dock the box onto the microscope to check if the microscope is ready for imaging; S24, open the EPU and select Atlas mode to get an overview of the ice distribution on the grid; move the sample stage to the area of ​​interest and acquire an image in GridSquare mode; find Concentric height, move the sample stage into the hole with appropriate ice layer and enter Hole / EucentricHeight mode; collect cryo-EM images in Data Acquisition mode; S25, if the carrier grid is not ideal and is not suitable for data collection, remove the carrier grid and repeat S24 to screen the next carrier grid; S26, after completing the screening of all automatic grids, if the ice thickness is not ideal, repeat S20-S25 to screen the grid, glow discharge time and Vitrobot parameters until the optimal ice thickness is obtained; if the RNA sample state is not suitable for data collection, screen the variables in in vitro transcription and RNA folding until the best micrograph suitable for data collection is observed; S27, activating the "ice filter" in the EPU to select holes based on the ice thickness determined by the image intensity; S28, Set up automatic data collection according to the EPU protocol and check the microscope and camera settings to ensure that data collection proceeds smoothly.

4. The method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA according to claim 2, characterized in that: S4 includes the following steps: S29, import all cryo-EM images and use MotionCor2 to perform "Motion Drift Correction" Correction", set "Number of patches X,Y" to "5,5", and set "Other MotionCor2 parameters" to "-Iter 10", the dose per frame is provided accordingly, and the rest of the parameters remain default; S30, perform "Contrast transfer function estimation" using CTFFIND4 with default settings, but set "Amount ofastigmatism" to Contrast ring out of The micrographs were visually inspected and selected; S31, using Blob picker in cryoSPARC, reference-based template matching in Relion, and reference-based CNN particle picking in EMAN2; S32, use the imported micrograph in S30 and the "input coordinates" generated in S31 to perform "particle extraction", and set an appropriate "particle frame size"; S33, perform "2D classification" on the scaled particles, set "Number of classes" to 200, "Use EMalgorithm" to "Yes", "Number of EM iterations" to 50, "Use VDAM algorithm" to "No", and "Mask diameter" to And set "Limit resolution E-step" to The rest of the parameters remain default; S34, visually inspect the classification results in "Subset selection" and discard those unwanted RNA particles and classes without obvious RNA features; S35, repeat S33-S34 2-3 times until most of the unnecessary particles are discarded; if no RNA feature is observed, terminate the data processing until the 2D classification results reveal the RNA feature and then perform data processing; S36, import particles from S35 and generate a "3D initial model", set "Number of classes" to 6, and keep the other parameters as default; S37, perform "3D classification" by importing all particles in S35 and a "reference image" in S36, set "Number of classes" to 6, and set "Limit resolution E-step" to The rest of the parameters remain default; S38, repeating S34 to select particles from the desired category, and then repeating S32 to extract the particles without scaling; S39, performing preliminary "3D automatic refinement" by importing the selected particles in S38 and a "reference image" in S37; S40, perform spectrum sharpening in "Post-processing" and import the "half unfiltered spectrum" and "solvent mask" in S39; S41, cryo-EM map visualized in Chimera; S42, repeat S37-S41 to obtain a better resolution of the flexible region density; S43, repeating steps S39-S41 for the selected category of particles and providing a corresponding "reference mask" to obtain a final structure that better resolves RNA heterogeneity; S44, import "CTF refinement", "grains" and "post-process STAR files" from S43, first "estimate magnification"; S45, import "Particles" from S44 into "CTF Refinement", perform "CTF Parameter Fitting", set "Fit Defocus" and "Fit B Factor" to "Per Particle", "Fit Astigmatism" to "Per Micrograph", "Esti mate beamtilt" and "Also estimate trefoil" to "Yes"; S46, import "micrographs" from S29, "particles" from S44, and "post-processed STAR files" from S43 into "Bayesian refinement", set "train optimal parameters" to "yes", set "perform particle refinement" to "no", and keep other parameters as default; S47, in "Bayesian Refinement", import the "Optimization Parameter File" from S46, set "Train Best Parameters" to "No", set "Perform Granular Refinement" to "Yes", and keep other parameters as default; S48, repeating S39-S41 to obtain a final structure from the refined particles; S49, analyze the "Local resolution" of the final map by providing the required input files and the "User-provided B-factor" from S48, as well as the corresponding "Calibrated pixel size", set "Use Relion" to "Yes", and import the "Detector MTF (STAR ​​file)" from the corresponding detector manufacturer; S50, Local resolution density map of S49 colored for presentation in Chimera.

5. The method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA according to claim 4, characterized in that: Using the Blob picker in cryoSPARC includes the following steps: S311, start a new "Blob Picker" job and import the micrographs from S30; S312, check "Show advanced" and set the expected minimum and maximum particle diameters. For Tetrahymena group I introns, this value is set to 50 and S313, open "Circular blob" and run the job, keeping the other parameters as default settings; S314, after completing the "Blob Picker" job, start a new "Inspect Particle Picks" job. Import the micrographs and particles from the "Blob Picker", keep the parameters at the default values, and adjust the "Power Histogram” to visually inspect the particle picking results for each micrograph; S315, when the selected particles are as many as possible and uniform, click "Done Picking".

6. The method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA according to claim 4, characterized in that: The use of reference-based template matching in Relion includes the following steps: S311, import the micrographs from S30 and set the corresponding "Pixel size in micrographs"; S312, import the results of the two-dimensional classification average, and set the "Lowpass filter references" to Set "Pixel size in references" as required.

7. The method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA according to claim 4, characterized in that: The use of reference-based CNN particle picking in EMAN2 includes the following steps: S311, navigate to "Particles->Interactive Particle Picking" in the GUI and import the micrograph from S30; interactively enter the "boxsize" and "ptclsize" values ​​determined from the micrograph. For Tetrahymena group I introns, these values ​​are set to 256 and 200. Fill in the remaining parameters; S312, manually select 15-20 reference images as good and background references; S313, train the CNN model with the "Threshold1" value set to 2; S314, click "Autobox" on a single micrograph to check the particle picking results; S315, when the selected particles are as many as possible and uniform, click "Autobox All".

8. The method for resolving RNA heterogeneity in determining the three-dimensional structure of RNA according to claim 4, characterized in that: In S39, import the "Reference mask" and set "Use solvent-flattened FSCs" to "Yes" to improve the reconstruction of low molecular weight and low signal-to-noise ratio particles.

9. A method for determining the three-dimensional structure of RNA, characterized in that The method comprises: RNA in vitro transcription and purification and RNA folding methods; The method for solving RNA heterogeneity according to any one of claims 1 to 8; and RNA modeling and methods for manual revision and new modeling based on RNA structure prediction.

10. An application of the method for determining the three-dimensional structure of RNA according to claim 9, characterized in that: Applicable to RNAs of 70-nt or greater that form three-dimensional structures.

Images