POINEER Platform

POI Network for Early Evaluation and Recognition

Analysis Workflow

The POINEER platform follows a four-step analysis pipeline:

graph LR A[MAF File] -->|Step 1| B[Basic Statistics
Mutation Spectrum
Sample Distribution] B -->|Step 2| C[Gene-Sample
Mutation Matrix] C -->|Step 3| D[Gene Set Mutation
Burden Matrix] D -->|Step 4| E[POI Risk Prediction
High/Low Risk] E -->|Step 5| F[Molecular Subtype
Distribution] style A fill:#e1f5ff style B fill:#fff4e1 style C fill:#e8f5e9 style D fill:#f3e5f5 style E fill:#ffebee style F fill:#e3f2fd
1

MAF File Analysis

Upload and analyze MAF format mutation files. Generate basic statistics, mutation spectrum plots, and sample distribution visualizations.

2

Gene-level Analysis

Convert site-level mutations to gene-level sample-gene mutation matrix. Count amino acid-changing mutations per gene per sample.

3

Gene Set Analysis

Calculate gene set mutation burden by counting mutated genes within each gene set per sample. Normalize by gene set size and perform statistical analysis.

Gene Set Mutation Burden Formula:
$$\mathrm{BURDEN}_{\text{term}} = \sum_{i} \left(\mathrm{Gene}_i \times \mathrm{Weight}_i\right)$$
4

Risk Prediction

Aggregate gene set mutations by cluster, calculate cluster-level mutation burden, and use pre-trained POINEER model to predict POI risk.

Cluster Mutation Burden Formula:
$$\mathrm{BURDEN}_{\text{cluster}} = \sum_{i} \mathrm{BURDEN}_{\text{term}_{i}}$$
5

Step 5: Molecular Subtype

Molecular subtyping for POI-predicted (high-risk) samples based on mutations in key genes, and distribution statistics.

Key Features

Easy File Upload

Support for MAF, CSV, TSV formats with intuitive drag-and-drop interface

Comprehensive Analysis

Complete workflow from mutation data to risk prediction with detailed statistics

AI-Powered Prediction

Pre-trained machine learning model for accurate POI risk assessment

Export Results

Download analysis results, matrices, and visualizations in multiple formats

Quick Start

  1. Navigate to Tools menu and select 1. MAF Analysis
  2. Upload your MAF format mutation file
  3. Follow the workflow through Steps 2, 3, and 4
  4. View and download your POI risk prediction results

MAF to Prediction Workflow

Complete analysis pipeline from MAF file to POI risk prediction

Analysis Workflow

The POINEER platform follows a four-step analysis pipeline:

1

MAF File Analysis

Upload and analyze MAF format mutation files. Generate basic statistics, mutation spectrum plots, and sample distribution visualizations.

2

Gene-level Analysis

Convert site-level mutations to gene-level sample-gene mutation matrix. Count amino acid-changing mutations per gene per sample.

3

Gene Set Analysis

Calculate gene set mutation burden by counting mutated genes within each gene set per sample. Normalize by gene set size and perform statistical analysis.

4

Risk Prediction

Aggregate gene set mutations by cluster, calculate cluster-level mutation burden, and use pre-trained POINEER model to predict POI risk.

5

Step 5: Molecular Subtype

Molecular subtyping for POI-predicted samples based on key gene mutations and subtype distribution statistics.

Step 1: Upload MAF File

Click to upload MAF file
Support MAF format files (.maf, .txt, .tsv)

MAF File Format

MAF (Mutation Annotation Format) files contain mutation data with columns such as Hugo_Symbol, Tumor_Sample_Barcode, Variant_Classification, Chromosome, etc.

  • Required columns: Hugo_Symbol, Tumor_Sample_Barcode, Variant_Classification
  • The analysis will generate basic statistics, mutation spectrum plots, and sample distribution visualizations
  • Results include mutation counts per sample, variant type distribution, and chromosome distribution

Step 1: MAF File Upload & Analysis

Upload and perform preliminary analysis on MAF format mutation files

Click to upload MAF file
Support MAF format files (.maf, .txt, .tsv)

MAF File Format

MAF (Mutation Annotation Format) files contain mutation data with columns such as Hugo_Symbol, Tumor_Sample_Barcode, Variant_Classification, Chromosome, etc.

  • Required columns: Hugo_Symbol, Tumor_Sample_Barcode, Variant_Classification
  • The analysis will generate basic statistics, mutation spectrum plots, and sample distribution visualizations
  • Results include mutation counts per sample, variant type distribution, and chromosome distribution

Step 2: Gene-level Mutation Analysis

Transform MAF mutations to gene-level sample-gene mutation matrix

Gene-level Analysis

This step converts site-level mutations from MAF files into a gene-level mutation matrix, counting amino acid-changing mutations per gene per sample.

  • Input: MAF file or gene-sample mutation matrix from previous step
  • Output: Sample × Gene mutation count matrix
  • Visualization: Mutation heatmap, gene mutation statistics, sample distribution
Click to upload MAF file or use previous results
Support MAF format or CSV matrix (samples × genes)

Analysis Parameters

Step 3: Gene Set Mutation Burden Analysis

Calculate gene set mutation burden and perform statistical analysis

Gene Set Analysis

This step calculates the number of mutated genes within each gene set per sample, normalizes by gene set size, and performs statistical analysis.

  • Input: Gene-sample mutation matrix from Step 2
  • Gene set data: GMT file or converted gene set definitions
  • Output: Sample × Gene Set mutation burden matrix with statistics
Click to upload gene-sample mutation matrix
CSV format: rows=samples, columns=genes

Analysis Parameters

Step 4: Disease Risk Prediction

Predict POI risk based on gene set cluster mutation burden

Risk Prediction

This step aggregates gene set mutations by cluster, calculates cluster-level mutation burden, and uses the pre-trained POINEER model to predict POI risk.

  • Input: Gene set mutation count matrix from Step 3
  • Cluster mapping: Excel file with gene set to cluster assignments
  • Model: Pre-trained MLP model (MCPmodel.pkl) for POI risk prediction
  • Output: Prediction probabilities, risk categories, and visualizations
Click to upload gene set mutation matrix
CSV format: rows=gene sets, columns=samples

Step 5: Molecular Subtype

Subtype POI-predicted samples based on key gene mutations

Molecular Subtype

For samples predicted as POI in Step 4, label protein K48-linked ubiquitination and/or positive regulation of mitotic nuclear division based on mutations in UBE3B/UBE4B and L3MBTL1/MAK/SPDL1. Then summarize counts and visualize the distribution.

  • Input: Gene mutation matrix (Step 2 output)
  • Input: Step 4 prediction results (CSV with sample_id and predict_label)
  • Output: Molecular subtype counts, barplot, and labeled prediction table
Upload Gene Matrix
CSV/TSV: rows=samples, columns=genes, values=mutation counts
Upload Prediction Results
CSV: includes sample_id and predict_label (from Step 4)

使用帮助

1. 平台简介

POINEER 基于样本遗传变异计算基因集突变负荷,并结合预训练模型评估原发性卵巢功能不全(POI)的发病风险。

注意:如需要使用模型进行预测,mutations应是MAF < 1% in EAS population in public databases (including ExAC, gnomAD, and 1000 Genomes)。

1.1 分析流程

POINEER平台提供四个步骤的完整分析流程:

  • 步骤1 - MAF文件分析:上传MAF格式突变文件,进行基础统计分析和可视化
  • 步骤2 - 基因水平分析:将位点突变转换为基因水平的样本-基因突变矩阵
  • 步骤3 - 基因集分析:计算基因集突变负荷,进行标准化和统计分析
  • 步骤4 - 风险预测:基于基因集cluster突变负荷,使用预训练模型预测POI风险
  • 步骤5 - 分子分型:根据关键基因突变对POI样本进行亚型标注并统计分布

2. 各步骤详细说明

2.1 步骤1 - MAF文件分析

功能:上传MAF格式突变文件,进行基础统计分析和可视化

输入格式:MAF (Mutation Annotation Format) 文件,必需列包括:

  • Hugo_Symbol:基因名称
  • Tumor_Sample_Barcode:样本ID
  • Variant_Classification:变异分类(如Missense_Mutation, Nonsense_Mutation等)
  • Chromosome:染色体位置(可选)

输出:基础统计信息、突变谱图、样本分布图、染色体分布图

2.2 步骤2 - 基因水平分析

功能:将MAF文件中的位点突变转换为基因水平的样本-基因突变矩阵

输入:MAF文件(来自步骤1)或已转换的基因-样本矩阵

输出格式:CSV文件,行为样本,列为基因,值为突变位点计数

输出内容:基因-样本突变矩阵、突变热图、基因突变统计

2.3 步骤3 - 基因集分析

功能:计算每个样本每个基因集中突变的基因个数,进行标准化和统计分析

输入:步骤2生成的基因-样本突变矩阵

基因集来源:GO Biological Process(c5.go.bp.v2025.1.Hs.symbols)。

基因集突变负荷计算公式

$$\mathrm{BURDEN}_{\text{term}} = \sum_{i} \left(\mathrm{Gene}_i \times \mathrm{Weight}_i\right)$$

公式说明:

  • $\mathrm{BURDEN}_{\text{term}}$ represents the mutational burden of a given BP term.
  • $\mathrm{Gene}_i$ denotes whether the i-th gene is mutated or not (coded as 1 for mutation while 0 for non-mutation).
  • $\mathrm{Weight}_i$ represents the occurrence frequency of the i-th gene within this BP term. Since a gene typically appears only once within a BP term, the weight is directly set to 1.

计算原理:对 BP term 内各基因的 Genei×Weighti 求和得到突变负荷;当 Weighti=1 时,该和等于该 term 内突变基因的数目。平台在可视化与统计中还可按基因集大小对负荷做归一化,以便比较不同大小的 term。

输出格式:CSV文件,行为基因集,列为样本,值为突变基因计数(标准化为百分比)

输出内容:基因集突变负荷矩阵、统计信息(均值、中位数、最大值、最小值、标准差)、Top基因集箱线图

2.4 步骤4 - 风险预测

功能:基于基因集cluster突变负荷,使用预训练模型预测POI风险

输入:步骤3生成的基因集突变计数矩阵

基因集Cluster突变负荷计算公式

$$\mathrm{BURDEN}_{\text{cluster}} = \sum_{i} \mathrm{BURDEN}_{\text{term}_{i}}$$

公式说明:

  • $\mathrm{BURDEN}_{\text{cluster}}$ represents the mutational burden of a given BP cluster.
  • $\mathrm{BURDEN}_{\text{term}_{i}}$ denotes score of the i-th BP term within this cluster.

计算原理:BP cluster 的突变负荷由该 cluster 内各 BP term 的 BURDENterm 得分按序求和得到;每个 term 的得分由步骤 3 的基因集(BP term)突变负荷公式计算。

处理流程:

  • 根据cluster映射文件,将基因集按cluster聚合
  • 计算每个cluster的突变负荷(使用加权求和公式)
  • 使用 GO2feature 映射表将矩阵特征名(Gene.set)转换为模型特征名(Features)
  • 使用预训练 MLP 模型(MCPmodel.pkl)进行POI风险预测

输出内容:预测概率(predict_prob)、预测标签(predict_label:1=高风险,0=低风险)、风险分布图、样本排名图

2.5 步骤5 - 分子分型

功能:在POI预测为高风险的样本中,根据UBE3B/UBE4B与L3MBTL1/MAK/SPDL1的突变状态进行分子亚型标注

输入:步骤2的基因突变矩阵 与步骤4的预测结果(predict_label)

输出:分子分型计数、分子分型barplot、包含分子亚型字段的预测结果表格与下载文件

3. 示例文件下载

您可以下载以下示例文件来了解正确的数据格式:

4. 分析参数说明

  • 步骤1(MAF分析):自动进行基础统计,无需额外参数
  • 步骤2(基因分析):可设置显示的前N个基因和样本数量(默认各50个)
  • 步骤3(基因集分析):可设置显示的前N个基因集(默认10个)
  • 步骤4(分类预测):使用预训练模型,预测阈值固定为0.5(predict_prob ≥ 0.5为高风险)
  • 步骤5(分子分型):无需额外参数,对Step 4预测为高风险的样本按关键基因突变状态进行分型

5. 结果解读

步骤4输出字段说明:

  • predict_prob:POI发病风险概率(0-1之间),数值越高表示风险越大
  • predict_label:预测标签,1表示高风险,0表示低风险
  • risk_category:风险类别,High Risk(高风险)或Low Risk(低风险)

预测阈值:默认使用0.5作为阈值,predict_prob ≥ 0.5预测为高风险,否则为低风险。

步骤5输出字段说明:

  • molecular_subtype:分子分型结果(可能同时包含两种亚型,若同时满足则用分号分隔)
  • subtype_protein_K48_ubiquitination:UBE3B/UBE4B突变标记(true/false)
  • subtype_positive_regulation_mitotic_nuclear_division:L3MBTL1/MAK/SPDL1突变标记(true/false)

6. 常见问题

MAF文件上传失败

请确保MAF文件包含必需的列:Hugo_Symbol, Tumor_Sample_Barcode, Variant_Classification。文件可以是制表符分隔(.maf, .tsv)或逗号分隔(.csv)格式。

基因矩阵格式错误

基因-样本矩阵应为CSV格式,第一行为样本ID,第一列为基因名称,矩阵值为突变计数(整数)。请参考示例文件格式。

分析时间过长

分析时间取决于数据大小:MAF文件分析通常需要几秒到几分钟;基因矩阵转换可能需要几分钟;基因集分析和风险预测通常较快(几秒到几十秒)。

如何理解预测结果?

predict_prob值在0-1之间,越接近1表示POI风险越高。默认阈值0.5:≥0.5为高风险,<0.5为低风险。您可以根据实际需求调整阈值。

7. 联系我们

如果您在使用本工具过程中遇到任何问题或有任何建议,请随时联系我们:

邮箱: support@ogi-tool.example.com
地址: Shanghai XuHui

我们的工作时间是周一至周五,9:00-17:00(北京时间)。我们会在24小时内回复您的邮件。

使用条款

1. 服务条款

欢迎使用POINEER指数计算平台。通过访问和使用本平台,您同意遵守以下使用条款和条件。如果您不同意这些条款,请不要使用本服务。

2. 数据隐私

我们重视您的数据隐私和安全:

  • 上传的数据仅用于POINEER指数计算,不会被存储或用于其他目的
  • 所有数据处理均在本地服务器完成,不会传输到第三方
  • 分析完成后,临时文件将被自动删除
  • 我们不会收集或存储任何个人身份信息

3. 使用限制

为确保服务质量和系统稳定,以下使用限制适用:

  • 单个文件大小不超过100MB
  • 每次分析的基因数量建议不超过50,000个
  • 请勿上传包含恶意代码或病毒的文件
  • 禁止进行任何可能损害系统安全的操作

4. 学术使用

本平台主要面向科研和学术用途:

  • 鼓励在学术研究中使用本平台进行数据分析
  • 如在发表的研究中使用了本平台,请适当引用
  • 商业用途需要事先获得许可

5. 免责声明

使用本平台时请注意:

  • 本平台提供的分析结果仅供参考,不构成医学或科学建议
  • 用户应对上传数据的合法性和准确性负责
  • 我们不对因使用本平台而产生的任何直接或间接损失承担责任
  • 平台可能因维护或升级而暂时不可用

6. 技术支持

如果您在使用过程中遇到技术问题或有任何疑问,请通过以下方式联系我们:

  • 查看帮助页面获取详细使用说明
  • 确保您的数据格式符合要求
  • 尝试使用示例数据验证平台功能

7. 条款更新

我们保留随时修改这些使用条款的权利。条款的任何更改将在本页面上发布。继续使用本平台即表示您接受修改后的条款。

最后更新时间:2026年03月11日