Author Archives: gene_x

单细胞RNA测序数据分析

scrna_seq_banner
scrna_seq_data_analysis

单细胞RNA测序使得细胞鉴定和研究在规模和分辨率上达到了高于批量测序的水平。

单细胞RNA测序(scRNA-seq)是分子生物学中发展和多样化最快的技术之一。研究基因表达在单个细胞水平上的能力就像之前批量RNA测序的出现一样具有变革性。

除了单细胞RNA测序,还有许多其他基于下一代测序(NGS)的检测方法已被适应于单细胞协议。这些包括基因组学、蛋白质组学和表观遗传学检测,特别是单细胞ATAC测序,通常与scRNA-seq一起进行。

平台和scRNA-seq协议在其吞吐量(细胞数)和转录本覆盖率(3’/5’标签基础 vs 全转录本)方面有所不同。我们团队在多种技术方面具有经验,如10X Genomics、Drop-Seq、BD Rhapsody系统以及CEL-Seq和Smart-Seq系列的协议。

这里我们介绍典型的单细胞分析,重点是scRNA-seq,但也涵盖了其与其他常见的单细胞检测的整合。

质量控制和预处理

与任何NGS数据一样,对单细胞测序数据的分析始于质量控制和预处理。

原始测序读数经过质量测试,并生成诸如细胞质量、准确性和多样性等指标。然后将读数与适当的参考基因组或转录组进行比对,并绘制和检查附加的指标,例如细胞数、每个细胞的读数、每个细胞的基因数、测序饱和度以及线粒体转录本的比例。

这些质控指标告诉我们关于文库的总体质量和样品的可用性,并使我们能够确定和去除低质量的细胞。

通常还进行进一步的预处理,以从某些下游分析中去除不需要的信号或噪声,这包括:

  • 估计替代值以估计由于技术而非生物学原因而导致的漏读或零转录本的基因;
  • 规范化以消除例如细胞大小差异等因素引起的偏差;以及
  • 将数据降至代表性变量,如高变量基因或主成分。
scrna_seq_qc
scrna_seq_imputation

探索性分析

预处理的单细胞RNA测序数据被聚类以识别相似的细胞群,并使用非线性降维算法(如tSNE和UMAP)和相关性热图进行可视化,以揭示细胞异质性的一般模式。

这些可视化帮助我们回答技术问题,例如:

  • 生物学重复是否相似?
  • 是否有离群样本或细胞?
  • 细胞群是否不同?

……以及生物学问题,例如:

  • 基础细胞类型/状态有多么异质?
  • 不同样本(例如不同组织、治疗或时间点)是否形成单独的群集?
scrna_seq_exploratory_analysis

细胞类型鉴定

识别和表征细胞类型(以及更精细的细胞状态)是大多数单细胞项目最核心的部分。

这一切始于识别特定于每个细胞群的特征(例如基因、蛋白质、可访问区域)。这些标记由差异表达(DE)比较每个细胞群和其余细胞群而定义,产生如折叠变化和统计显着性等DE统计量。

可以使用散点图、小提琴图和热图可视化细胞群标记。

标记进一步注释为生物学意义的术语,例如生物过程、信号通路或特定疾病。这些分析可能依赖于超表达分析或基因集富集分析,两者都会产生一系列富集的基因集与相关统计信息和注释。

单细胞数据集通常也与公共可用数据集集成,以利用已注释数据集或细胞图谱中的细胞类型信息。这使得将细胞标签转移至分析的数据集成为可能。

转移的细胞标签和鉴定的标记及其注释与关于细胞类型/状态标记的先前信息一起用于鉴定捕获的细胞类型。

scrna_seq_marker_detection
scrna_seq_marker_visualization

轨迹分析

除了表征不同的细胞身份外,单细胞数据还适用于识别细胞状态渐变的连续体或轨迹。揭示这种连续体也被称为假时间分析,尽管所有细胞在同一时间点被采样,但个体细胞可能代表分化等时间过程中不同的阶段。

利用分化分支和细胞成熟轨迹的全新重建,可以探索细胞动态,勾勒细胞发育谱系,并表征沿着潜在假时间维度的细胞状态转换。

轨迹推断算法的集合可用于鲁棒地识别根和终端细胞状态、分支点和谱系。单细胞沿着确定性或概率谱系进行排序,它们的排序指示了它们在感兴趣的动态过程中的进展情况。

这种类型的分析还可以利用加工和未加工转录本的比率推断基因表达在给定细胞中是增加还是减少。将来自给定状态下所有定量基因的这些信息相结合,可以推断状态的变化方向和速度。这称为RNA速度分析。

scrna_seq_trajectory_analysis

综合单细胞分析

综合单细胞分析将不同的数据集,包括不同的数据类型和物种集成在一起,这使得对所研究系统中基因调控的机制有更准确和详细的细胞标记和洞察。这种分析依赖于数据集之间的共同属性或“锚点”,如匹配的特征(例如基因或同源物)或匹配的细胞。

整合多个单细胞RNA测序数据集

最常见的单细胞数据集整合是来自不同来源或技术平台的scRNA-seq数据集之间的整合。使用基因作为锚点,成功的整合可以去除数据集的技术偏差同时保留生物变异。

当有关于相应组织或生物体的公共表达图谱时,整合不同的scRNA-seq数据集特别有帮助。

整合单细胞RNA测序和表观遗传学

将单细胞RNA测序数据与单细胞ATAC-seq或单细胞甲基化数据结合起来通常依赖于匹配的细胞作为锚点(当测量来源于与10X Genomics Multiome技术中相同的细胞时)。

将表达数据与染色质可及性或甲基化数据相结合,可以更可靠地识别细胞类型,并允许量化染色质状态对各个细胞类型的表达的影响。

阅读有关整合表观遗传学和转录组学的更多信息

整合单细胞RNA测序和蛋白质组学

由于蛋白质而不是转录本是细胞功能的关键驱动因素,单细胞蛋白质组学通过更准确地估计细胞的功能状态来补充scRNA-seq实验。

单细胞蛋白质组学分析(CITE-seq,流式细胞术,质谱和质谱分析)具有不同的吞吐量(量化的蛋白质数量)并可以专门针对表面蛋白进行定向,如CITE-seq,它涉及从具有匹配scRNA-seq读数的细胞中量化表面蛋白。

表面蛋白在细胞类型鉴定中特别有用,而包含胞质蛋白则可以更好地表征通路和基因调控活动。

跨物种整合分析

跨物种综合分析可确定定义不同生物之间进化和发育机制关系的细胞类型谱系。在跨物种整合中,使用共享的同源物作为锚点。

当疾病/器官在动物模型中的单细胞分辨率上得到更好的表征时,这特别有助于人类疾病/器官的研究。

scrna_seq_integrative_analysis

膜受体配体分析

膜受体配体(LR)分析揭示了协调体内稳态、发育和其他系统级功能的细胞间相互作用。此类相互作用的变化和功能失调在仅限于个体细胞或细胞类型内部状态分析中可能不被注意到。

膜受体配体分析根据已知受体和其配体的表达量识别和量化细胞间相互作用。这些相互作用可能在组织内或组织间发生,其强度将在感兴趣的生物条件(如患者组、疾病状态和治疗)之间进行比较。

scrna_seq_ligand_receptor_analysis

空间转录组分析

空间解析单细胞转录组学分析将表达数据与细胞在组织或器官中的位置上下文联系起来。这在研究肿瘤及其微环境等复杂实体组织中特别有用。

空间转录组分析包括空间中的细胞/点聚类、空间变量基因的识别和空间中的细胞类型解析。

保留序列化细胞的位置信息有助于准确识别细胞类型和膜受体配体相互作用。它还能够实现基因表达或染色质可及性(在scATAC-seq情况下)的空间可视化,并将基于成像的数据整合到分析中。

即使在像10X Visium这样的低分辨率分析中,多模式空间分析也有助于纠正基因表达值和补充数据缺失事件。

scrna_seq_spatial_analysis

RNA测序数据分析

rna_seq_banner6
rna_seq_data_analysis

RNA测序数据分析揭示了基因调控的复杂机制。

基因表达的转录组广泛应用于研究从 单细胞 到组织和复杂的微生物群落中的生物系统中的基因调控的研究。 RNA测序数据允许进行各种分析,以解决生物学和生物医学领域中无数的研究问题。

下面我们介绍了我们在RNA-seq数据上执行的一些最常见的分析。探索性、差异表达和通路分析大多也适用于其他高通量表达数据,如表达型芯片或蛋白质组学数据。

我们希望下面的示例能启发您欣赏RNA-测序的丰富多彩世界。

探索性基因表达分析

每个RNA-seq表达研究都包括探索性分析。在经过原始测序reads质量控制和基因计数之后,使用主成分分析(PCA)和表达热图来可视化数据集,以揭示其一般模式。这些可视化帮助我们回答以下问题:

  • 生物学重复是否与其表达剖面相似?
  • 不同样本组(例如不同组织、处理或时间点)是否形成单独的聚类?
  • 是否存在异常样本?
rna_seq_exploratory_analysis

差异表达分析

差异表达分析是对两个样本组进行统计比较的过程。它会得到每个检测到的转录本的差异表达统计数据,例如折叠差异和统计显著性。这些统计数据通常使用火山图进行可视化。被发现上调或下调的基因可以进一步通过热图或箱线图进行可视化。

作为一种统计分析方法,表达研究中的这个阶段受益于生物复制品带来的统计功率。每个条件至少需要三个生物重复样本,但这仅适用于可靠检测具有相对较大表达差异的基因。通过谨慎的实验设计和足够的样本量,可以检测到更微妙的差异,并控制混杂因素。

rna_seq_differential_expression_analysis

通路分析

通路分析将差异表达分析中的基因放在更广泛的生物学背景中。简单的通路分析会将上调和下调基因与预定的基因列表进行统计学比较。这些列表被注释为生物学意义的术语,例如生物过程、信号通路或特定疾病。

这样的分析可能依靠过表达分析或基因集富集分析,两者都会得出具有相关统计学和注释的富集基因集列表。

更多机制通路分析依赖于基因之间实验验证的相互作用。它们不仅能够确定哪些通路由差异表达的基因表示,还能揭示通路是否被激活或抑制,以及由哪些基因激活或抑制。

更高级的通路分析我们使用Ingenuity Pathway Analysis (IPA, QIAGEN)。IPA能够进行深入分析已知和新颖的基因调控网络。

rna_seq_pathway_analysis.996x0-is

转录组组装

对于非模式生物以及具有非常动态的基因组,例如微生物,我们通常通过组装新的转录组来开始RNA测序数据分析,并使用相关物种的同源基因和计算基因预测来注释它。

一个新的参考转录组对您的进一步研究和整个研究社区的研究都是非常宝贵的资源。一旦建立了高质量的参考转录组,就可以打开大多数下游分析的大门,这些下游分析通常用于模型生物。

rna_seq_transcriptome_assembly

单细胞表达分析

单细胞RNA测序(scRNA-seq)实验可以以比批量RNA测序更高的规模和分辨率对细胞类型进行编目和揭示分化轨迹。

特别是用于研究复杂组织的组成和发展,scRNA-seq数据集通常包含数千个单个细胞。大多数用于分析批量RNA-seq数据的方法也可以为单细胞RNA-seq数据量身定制。

了解更多

scrna_seq_analysis

MicroRNA数据分析

小RNA测序可用于研究各种短RNA物种,尤其是microRNAs。MicroRNA-seq分析与mRNAs的分析主要类似,但路径和调节分析利用预测和/或先前验证过的microRNA靶基因。

从匹配样品中同时测序mRNA和小RNA可估计microRNAs与其靶标之间的调节关系。为了确定在给定条件下受microRNA调节的基因,可以使用argo naute CLIP-测序(和相关协议)。

mirna_seq_analysis

可变剪接 分析

除了在基因水平上研究表达外,RNA测序还允许进行更详细的视图:剪接变异水平的表达。可靠地鉴定可变剪接事件需要比典型的基因水平表达分析更深的测序。

根据数据的数量和质量,可变剪接分析可以集中于量化已知的、先前注释的剪接亚型的表达水平,或检测新的剪接事件。

rna_seq_alternative_splicing_analysis

融合基因检测

在癌症中,某些结构变异已知会导致融合基因。DNA中两个分开的基因融合在一起可能导致融合转录本。反过来,融合转录本可能导致融合蛋白质具有新的、潜在的癌症驱动调控和功能组合。

可以使用识别和分析discordantly mapping RNA-seq读数或读取对的工具从RNA-seq数据中检测融合基因。

rna_seq_fusion_detection

整合RNA-seq和表观基因组数据

在同一样本上进行RNA-seq和表观基因组测序(例如ChIP或ATAC-seq)可以进行整合分析,研究基因调控程序的全基因组范围。

可以在基因表达和调控元素的表观基因组状态的证据基础上,确定增强子与其靶基因以及转录因子与其靶基因之间的调控联系。

了解更多

screen_shot_2022-11-04_at_4_31_31_pm

微生物与病毒生物信息学 vs. 细菌与病毒生物信息学

核心区别

  • Microbial & Viral Bioinformatics(微生物与病毒生物信息学)
    研究对象更广,涵盖所有类型的微生物(细菌、古菌、真菌、原生生物等)以及病毒。
  • Bacterial and Viral Bioinformatics(细菌与病毒生物信息学)
    研究对象较窄,专注于细菌和病毒,不包括其他非细菌类微生物。
Bacterial_and_Viral_Bioinformatics

对比表

维度 微生物与病毒生物信息学 细菌与病毒生物信息学
研究对象范围 微生物(细菌、古菌、真菌、原生生物等)+ 病毒 仅限细菌 + 病毒
广度 广,涵盖多类生物 窄,专注于细菌相关
典型研究方向 宏基因组学、群落生态学、宿主-微生物互作、病毒-微生物关系 细菌基因组学、耐药基因预测、病原体分析、噬菌体研究
应用场景 微生物群落研究、环境样本分析、人体微生物组研究 医学细菌学、传染病、抗生素耐药机制
学科定位 偏向综合性、生态学与系统层面 偏向临床、病原学与应用层面

应用实例对比

  • 微生物与病毒生物信息学

    • 实例:分析人类肠道微生物组(细菌、真菌、古菌等)与病毒群落的互作,探索其与肥胖、糖尿病或免疫系统疾病的关系。
    • 特点:关注整体微生物生态系统,强调多类生物之间的协同与平衡。
  • 细菌与病毒生物信息学

    • 实例:研究医院获得性耐药细菌(如耐甲氧西林金黄色葡萄球菌,MRSA)及其与噬菌体的互作,寻找新的治疗策略。
    • 特点:聚焦病原体与临床应用,强调对疾病防控和治疗的直接价值。

总结

  • Microbial & Viral Bioinformatics:范围大,适合研究多样微生物群落及其与病毒的关系。
  • Bacterial and Viral Bioinformatics:范围小,专注于细菌和病毒,更聚焦临床和病原学研究。

import matplotlib.pyplot as plt
from matplotlib_venn import venn2

# Create side-by-side comparison
fig, axes = plt.subplots(1, 2, figsize=(12,6))

# --- Left: Microbial & Viral ---
venn_left = venn2(
    subsets=(1, 1, 1),
    set_labels=("Microbes", "Viruses"),
    ax=axes[0],
    set_colors=("skyblue", "lightcoral"),  # 微生物用蓝色,病毒用红色
    alpha=0.6
)
venn_left.get_label_by_id('10').set_text('Bacteria, Archaea,\nFungi, Protists')
venn_left.get_label_by_id('01').set_text('Viruses')
venn_left.get_label_by_id('11').set_text('Microbe-Virus\ninteractions')
axes[0].set_title("Microbial & Viral Bioinformatics")

# --- Right: Bacterial & Viral ---
venn_right = venn2(
    subsets=(1, 1, 1),
    set_labels=("Bacteria", "Viruses"),
    ax=axes[1],
    set_colors=("lightgreen", "lightcoral"),  # 细菌用绿色,病毒用红色
    alpha=0.6
)
venn_right.get_label_by_id('10').set_text('Bacteria\n(pathogens, commensals)')
venn_right.get_label_by_id('01').set_text('Viruses\n(phages, human viruses)')
venn_right.get_label_by_id('11').set_text('Bacteria-Virus\ninteractions')
axes[1].set_title("Bacterial & Viral Bioinformatics")

# Add a caption below the plots
fig.text(
    0.5, -0.05,
    "Comparison: Microbial & Viral Bioinformatics covers all microbes (bacteria, archaea, fungi, protists) plus viruses,\n"
    "while Bacterial & Viral Bioinformatics is narrower, focusing only on bacteria and viruses.",
    ha='center', fontsize=10
)

plt.tight_layout()
plt.show()
import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(6,6))

# 大圆 - Microbial
microbe_circle = plt.Circle((0,0), 1.0, color="skyblue", alpha=0.4, label="Microbes")
ax.add_artist(microbe_circle)

# 小圆 - Bacteria (在Microbial里)
bacteria_circle = plt.Circle((0.3,0.3), 0.4, color="lightgreen", alpha=0.6, label="Bacteria")
ax.add_artist(bacteria_circle)

# 另一个圆 - Viruses
virus_circle = plt.Circle((-0.2,-0.2), 0.6, color="lightcoral", alpha=0.6, label="Viruses")
ax.add_artist(virus_circle)

# 设置比例 & 美化
ax.set_xlim(-1.5, 1.5)
ax.set_ylim(-1.5, 1.5)
ax.set_aspect("equal")
ax.axis("off")

# 添加标签
ax.text(-0.9, 1.0, "Microbes\n(bacteria, archaea, fungi, protists)", fontsize=10, color="blue")
ax.text(0.4, 0.6, "Bacteria", fontsize=10, color="green")
ax.text(-0.9, -0.8, "Viruses", fontsize=10, color="red")

plt.title("Nested Scope: Microbes vs Bacteria with Viruses", fontsize=12)
plt.show()

Analyse spatial transcriptomic data

Click here (http://xgenes.com/course/spatialtx/) to get started.

Analyzing spatial transcriptomic data involves several steps, including quality control, data pre-processing, spatial mapping, cell-type identification, differential expression analysis, and functional analysis. Here are the main steps involved in spatial transcriptomic data analysis:

Quality control: This step involves checking the quality of the sequencing reads using tools such as FastQC. If the quality is low, the data may need to be re-sequenced or filtered to remove low-quality reads or adapter sequences. Data pre-processing: This step involves filtering out low-quality cells, normalizing the data, and identifying highly variable genes using tools such as ST Pipeline or STARmap. Spatial mapping: This step involves mapping the transcriptomic data to the spatial coordinates of the tissue sections using tools such as ST Pipeline or STARmap. This produces a spatial expression matrix that contains the expression levels of each gene in each spatial location. Cell-type identification: This step involves identifying the cell types that are present in each spatial location using tools such as CellFinder or SpatialDE. Differential expression analysis: This step involves identifying genes that are differentially expressed between different spatial locations or cell types. This can be done using tools such as ST Pipeline or SpatialDE. Functional analysis: This step involves interpreting the differentially expressed genes by performing pathway or gene ontology analysis. This can be done using tools such as GSEA or Enrichr. Data visualization: This step involves visualizing the results of the analysis using tools such as heatmaps, spatial plots, or 3D visualizations. Overall, analyzing spatial transcriptomic data is a complex process that involves several steps and tools. It is important to carefully QC the data, choose appropriate normalization and statistical methods, and interpret the results in the context of the biological question being studied. Additionally, there are specialized tools and methods available for different types of spatial transcriptomic data, such as MERFISH or CODEX, which may require different analysis pipelines.

Anlalyse single cell transcriptomic data

Click here (http://xgenes.com/course/singlecelltx/) to get started.

Analyzing single-cell transcriptomic data involves several steps, including quality control, data pre-processing, cell clustering, differential expression analysis, and functional analysis. Here are the main steps involved in single-cell transcriptomic data analysis:

Quality control: This step involves checking the quality of the sequencing reads using tools such as FastQC. If the quality is low, the data may need to be re-sequenced or filtered to remove low-quality reads or adapter sequences. Data pre-processing: This step involves filtering out low-quality cells, normalizing the data, and identifying highly variable genes using tools such as Seurat or Scanpy. Cell clustering: This step involves grouping cells that have similar gene expression profiles into clusters using unsupervised clustering algorithms such as k-means or hierarchical clustering. This can be done using tools such as Seurat, Scanpy, or RaceID. Differential expression analysis: This step involves identifying genes that are differentially expressed between different clusters of cells. This can be done using tools such as Seurat or Scanpy. Functional analysis: This step involves interpreting the differentially expressed genes by performing pathway or gene ontology analysis. This can be done using tools such as GSEA or Enrichr. Data visualization: This step involves visualizing the results of the analysis using tools such as t-SNE, UMAP, or heatmaps. Overall, analyzing single-cell transcriptomic data is a complex process that involves several steps and tools. It is important to carefully QC the data, choose appropriate normalization and statistical methods, and interpret the results in the context of the biological question being studied. Additionally, there are specialized tools and methods available for different types of single-cell transcriptomic data, such as scRNA-seq, scATAC-seq, or spatial transcriptomics, which may require different analysis pipelines.

Anlalyse RNA-seq data

Click here (http://xgenes.com/course/rnaseq/) to get started.

Analyzing RNA-seq data involves several steps, including quality control, alignment or mapping of the reads to a reference genome or transcriptome, quantification of gene expression, normalization, differential gene expression analysis, and functional analysis of the differentially expressed genes. Here are the main steps involved in RNA-seq data analysis:

Quality control: This step involves checking the quality of the sequencing reads using tools such as FastQC. If the quality is low, the data may need to be re-sequenced or trimmed to remove low-quality reads or adapter sequences. Alignment or mapping: This step involves aligning the sequencing reads to a reference genome or transcriptome using tools such as HISAT2, STAR, or Tophat. This step produces alignment files (in BAM or SAM format) that are used in the next step. Quantification: This step involves quantifying gene expression levels from the aligned reads using tools such as featureCounts or HTSeq. This produces a count matrix that contains the number of reads mapped to each gene. Normalization: This step involves normalizing the count matrix to account for differences in sequencing depth or library size between samples. Common normalization methods include TMM or FPKM. Differential gene expression analysis: This step involves identifying genes that are differentially expressed between two or more groups of samples. This is typically done using statistical tests such as the Wald test or the Likelihood Ratio test in packages such as DESeq2, edgeR, or limma-voom. Functional analysis: This step involves interpreting the differentially expressed genes by performing pathway or gene ontology analysis. This can be done using tools such as DAVID, Enrichr, or GSEA. Overall, analyzing RNA-seq data is a complex process that involves several steps and tools. It is important to carefully QC the data, choose appropriate normalization and statistical methods, and interpret the results in the context of the biological question being studied.

Explore microbiome data with Phyloseq

Click here (http://xgenes.com/course/phyloseq/) and Click here (http://xgenes.com/course/phyloseq2/) to get started.

Phyloseq is an R package for the analysis of microbiome data. Microbiome data is generated from high-throughput sequencing technologies such as 16S rRNA gene sequencing or metagenomic sequencing, which allow for the identification and quantification of the microbial species present in a given sample.

Phyloseq provides a framework for importing, analyzing, and visualizing microbiome data in R. It allows users to perform a wide range of analyses, including alpha and beta diversity analyses, differential abundance testing, and network analysis. Additionally, Phyloseq integrates with other R packages for statistical analysis and data visualization.

Overall, Phyloseq provides a powerful toolset for exploring microbiome data and generating insights into the microbial communities present in a given sample.

Align Sequences with MUSCLE

Click here (http://sishi.com/muscleform) to get started.

MUSCLE (Multiple Sequence Comparison by Log-Expectation) is a software program used for multiple sequence alignment of nucleotide or protein sequences. It is a fast and efficient program that can align large numbers of sequences with high accuracy. It was developed by Robert C. Edgar, who first described it in a paper titled “MUSCLE: multiple sequence alignment with high accuracy and high throughput” published in the journal Nucleic Acids Research in 2004.

Optimize your RNA secondary structure with RNAHeliCes

Click here (http://sishi.com/rnahelicesform) to get started.

RNAHeliCes is a computational tool used to analyze RNA folding space. RNA molecules can fold into a variety of structures, and understanding the folding process is important for studying their biological functions. RNAHeliCes uses a simplified representation of RNA structures, in which each structure is abstracted using position-specific helices. This approach allows RNAHeliCes to analyze large RNA datasets efficiently, by reducing the complexity of the structures and focusing on their most important features.

The method behind RNAHeliCes was developed by a team of researchers led by Jiabin Huang, Rolf Backofen, and Björn Voß. The work was published in the scientific journal RNA in 2012, and the paper is titled “Abstract folding space analysis based on helices.” The paper describes the method in detail, including how the helix-based abstraction is implemented and how it can be used to analyze RNA folding space. We also provide examples of how RNAHeliCes can be used to analyze RNA molecules with known structures, and how it can be used to predict the structures of RNA molecules with unknown structures.