GEO2R差异表达分析参数如何设置？2026实操指南
AIAI Summary (BLUF)
This content describes the options available in GEO2R for differential expression analysis, including P-value adjustment methods (Benjamini & Hochberg FDR, Bonferroni, etc.), log transformation, limma
Introduction
The GEO2R web tool provides a user-friendly interface for comparing two or more groups of samples in GEO datasets to identify differentially expressed genes. Before performing an analysis, users must configure several key parameters that control statistical testing, data transformation, normalization归一化：将数据标准化以消除系统性偏差。, and visualization. This blog post explains each option in detail, helping you make informed choices for reproducible and robust results.
GEO2R 网络工具提供了一个用户友好的界面，用于比较 GEO 数据集中两个或多个样本组，以识别差异表达基因。在执行分析之前，用户必须配置多个关键参数，这些参数控制统计检验、数据变换、归一化和可视化。本博客文章详细解释了每个选项，帮助您做出明智的选择，以获得可重复且稳健的结果。
Key Options for Differential Expression Analysis
P‑value Adjustment Methods
When performing multiple hypothesis tests, raw p‑values must be adjusted to control the rate of false positives. GEO2R offers six correction methods, each with different statistical properties. The table below summarizes their characteristics and typical use cases.
在执行多重假设检验时，必须对原始 p 值进行校正以控制假阳性率。GEO2R 提供了六种校正方法，每种方法具有不同的统计特性。下表总结了它们的特征和典型使用场景。
| Method | Description | Use Case |
|---|---|---|
| Benjamini & Hochberg (False discovery rate)Benjamini-Hochberg错误发现率：一种控制假阳性比例的方法。 | Controls the expected proportion of false positives among rejected hypotheses. | Default choice for most genomic studies; balances discovery and error control. |
| Benjamini & Yekutieli | A more conservative FDR method that assumes dependency among tests. | Use when tests are known to be positively or arbitrarily dependent. |
| Bonferroni | Divides α by the number of tests; very strict control of family‑wise error rate (FWER). | Strong correction, suitable when any false positive is unacceptable (e.g., clinical markers). |
| Hochberg | A step‑up procedure that controls FWER under certain conditions. | Less conservative than Bonferroni but still controls FWER; assumes positively correlated tests. |
| Holm | A step‑down Bonferroni‑based procedure that controls FWER uniformly. | Often preferred over Bonferroni because it is less conservative while maintaining strong FWER control. |
| Hommel | A step‑up procedure that controls FWER; uniformly more powerful than Hochberg when tests are positively dependent. | Use when tests are positively correlated and a powerful FWER control is desired. |
Recommendation: For most exploratory analyses, select Benjamini & Hochberg (FDR). If you need strict control of any false positive, choose Bonferroni or Holm.
建议：对于大多数探索性分析，选择Benjamini & Hochberg (FDR)。如果需要严格控制任何假阳性，请选择Bonferroni或Holm。
Log TransformationLog转换：对数据进行对数变换，通常用于正态化分布。 and Normalization归一化：将数据标准化以消除系统性偏差。
Raw expression data often require log transformationLog转换：对数据进行对数变换，通常用于正态化分布。 to stabilize variance and make the data more normally distributed. GEO2R provides three log transformationLog转换：对数据进行对数变换，通常用于正态化分布。 options:
- Auto‑detect: The tool examines the data and automatically applies log2 transformation if the data appear not to be log‑transformed.
- Yes: Force log2 transformation regardless of the current state.
- No: Do not apply any log transformationLog转换：对数据进行对数变换，通常用于正态化分布。.
自动检测：工具检查数据，如果数据看起来未经过对数变换，则自动应用 log2 变换。
是：强制进行 log2 变换，无论当前状态如何。
否：不应用任何对数变换。
Additionally, the limma precision weights (vooma)limma精度权重（vooma）：估计基因特异性权重以处理异方差性的方法。 option applies a variance‑modelling approach that adjusts for the mean‑variance relationship often seen in microarray or count data. Enable this when you expect heteroscedasticity across expression levels.
此外，limma 精确权重 (vooma) 选项应用了一种方差建模方法，调整微阵列或计数数据中常见的均值‑方差关系。当您预期不同表达水平存在异方差性时，启用此选项。
The Force normalization归一化：将数据标准化以消除系统性偏差。 option forces quantile normalization归一化：将数据标准化以消除系统性偏差。 of the expression matrix. Normalization归一化：将数据标准化以消除系统性偏差。 is critical for removing technical variation between samples. Use this when samples are not already normalized or when you want to ensure consistent processing.
强制归一化选项强制对表达矩阵进行分位数归一化。归一化对于消除样本间的技术变异至关重要。当样本尚未归一化或您希望确保处理一致性时，使用此选项。
| Option | Description | When to Use |
|---|---|---|
| Auto‑detect log | Automatic log2 transformation based on data range. | Safe default; works well for most microarray and RNA‑seq. |
| Yes (log) | Force log2 transformation. | When auto‑detect fails or you want explicit control. |
| No (log) | No log transformationLog转换：对数据进行对数变换，通常用于正态化分布。. | Data already log‑transformed or for special analyses (e.g., linear models on raw counts). |
| vooma | Apply limma precision weights. | For datasets with non‑constant variance (e.g., RNA‑seq with low counts). |
| Force normalization归一化：将数据标准化以消除系统性偏差。 | Quantile normalize expression matrix. | Samples were processed in different batches or show strong technical artifacts. |
Configuring Plots and Contrasts
After setting statistical options, you can customize the visualization parameters. Two critical inputs are the Significance level cut‑off (adjusted p‑value threshold) and the Log2 fold changeLog2倍数变化：以2为底的对数折叠变化，用于量化表达差异。 threshold. These define which genes are considered significantly differentially expressed in the volcano and mean‑difference (MA) plots.
设置统计选项后，您可以自定义可视化参数。两个关键输入是显著性水平截断值（校正后 p 值阈值）和Log2 倍数变化阈值。这些定义了在火山图和均值差异 (MA) 图中哪些基因被视为显著差异表达。
- Significance level cut‑off: Enter a number between 0 and 1 (e.g., 0.05). Genes with adjusted p‑value below this value are marked as significant.
- Log2 fold changeLog2倍数变化：以2为底的对数折叠变化，用于量化表达差异。 threshold: A positive number (e.g., 1.0 means 2‑fold change). Genes with absolute log2 fold changeLog2倍数变化：以2为底的对数折叠变化，用于量化表达差异。 above this threshold are considered biologically meaningful.
显著性水平截断值：输入 0 到 1 之间的数字（例如 0.05）。校正后 p 值低于此值的基因被标记为显著。
Log2 倍数变化阈值：一个正数（例如 1.0 表示 2 倍变化）。绝对 log2 倍数变化高于此阈值的基因被认为具有生物学意义。
Contrasts: The volcano and MA plots require at least two groups to be defined. After defining sample groups, select up to five contrasts (pairwise comparisons) to display. The contrast selection applies only to plot generation; the full differential expression table will still contain results for all possible comparisons.
对比：火山图和 MA 图需要至少定义两个组。定义样本组后，选择最多五个对比（两两比较）进行显示。对比选择仅适用于图生成；完整的差异表达表仍将包含所有可能比较的结果。
Running the Analysis
Once all options are configured, click the Reanalyze button to re‑run the differential expression analysis with the updated settings. If you change options after an initial analysis, a new analysis will be triggered. Note that distribution plots can be viewed even without specifying sample groups.
配置所有选项后，点击重新分析按钮，使用更新后的设置重新运行差异表达分析。如果在初始分析后更改选项，将触发新的分析。注意，即使未指定样本组，也可以查看分布图。
For further technical details, refer to the official GEO2R documentation linked in the interface.
更多技术细节，请参考界面中链接的官方 GEO2R 文档。
常见问题（FAQ）
GEO2R中P值校正方法怎么选？
探索性分析选Benjamini & Hochberg (FDR)；需严格控制假阳性选Bonferroni或Holm。
GEO2R的log变换和归一化选项怎么设置？
通常选自动检测log2变换；若样本未归一化，启用强制归一化（分位数归一化）。
GEO2R中vooma权重是什么？什么时候用？
vooma是limma精度权重，校正表达水平的异方差性。当数据存在均值-方差关系时启用。
版权与免责声明：本文仅用于信息分享与交流，不构成任何形式的法律、投资、医疗或其他专业建议，也不构成对任何结果的承诺或保证。
文中提及的商标、品牌、Logo、产品名称及相关图片/素材，其权利归各自合法权利人所有。本站内容可能基于公开资料整理，亦可能使用 AI 辅助生成或润色；我们尽力确保准确与合规，但不保证完整性、时效性与适用性，请读者自行甄别并以官方信息为准。
若本文内容或素材涉嫌侵权、隐私不当或存在错误，请相关权利人/当事人联系本站，我们将及时核实并采取删除、修正或下架等处理措施。 也请勿在评论或联系信息中提交身份证号、手机号、住址等个人敏感信息。