← Baciptosis Subtype Labv1.0.0

使用说明 / User guide

所有计算在您的浏览器内完成 / All analysis runs locally in your browser.

RNA 表达 / RNA expression

CSV/TSV第一列为基因符号,其余每列为一个患者。必须包含BAX、BAK1、VDAC1、VDAC2、PHB1、PHB2。PHB1缺失时才使用PHB。明确选择TPM或log2(TPM+1)。缺失、重复及负表达不接受。

First column: gene symbol. Remaining columns: unique patient IDs. Include all six genes. PHB is used only if PHB1 is absent. Select TPM or log2(TPM+1). Missing values, duplicate target genes, and negative expression are rejected. Raw counts, FPKM, FASTQ and qPCR are not supported.

固定TCGA均值/总体标准差及中心用于单例和批量赋类,不按新队列重新归一化。输出为研究性最近中心赋类,不是原共识聚类的完全等价实现。TCGA一致率485/516;CPTAC固定参考119/120被赋为C1,提示显著平台偏移。目前不适合直接用于临床分型。

The frozen TCGA population mean/SD and centroids are used for both single and batch cases. This is a research nearest-centroid surrogate. TCGA concordance: 485/516. With fixed-reference normalization, 119/120 CPTAC tumors were assigned C1, indicating substantial platform shift. This is not validated clinical subtyping. Centroid distances are not probabilities.

UNI2-h 病理特征 / Pathology features

H5支持N×1536或1×N×1536。每个文件是一个切片,请检查患者编号和数据集;同患者多切片填写相同患者编号。使用训练时相同的UNI2-h编码器及肿瘤区域/patch预处理。不会上传或分析原始H&E图片。

H5 supports N×1536 or 1×N×1536, one slide per file. Check the patient mapping and selected dataset. Assign the same patient ID to slides from one patient. Use the same UNI2-h encoder and tumor/patch preprocessing as training. Original H&E images are not accepted.

均值CSV/TSV表头:patient_id,可选slide_id,feature_0001至feature_1536。含slide_id时按患者等权平均切片;不含slide_id时每行必须是一个唯一患者的均值向量。维度顺序必须与UNI2-h原始输出一致。

Mean-feature CSV/TSV: patient_id, optional slide_id, feature_0001 … feature_1536 in order. With slide_id, slide means are equally averaged per patient. Without slide_id, each row must contain one unique patient's mean vector. Feature order must match UNI2-h output.

超大文件或压缩不兼容时,可下载convert_h5.py在本地转换。需要numpy、h5py。清单列为patient_id,slide_id,path,可选dataset。命令:python convert_h5.py --manifest slides.csv --output means.csv。

For large or unsupported H5 files, run the local converter with numpy and h5py. Manifest columns: patient_id,slide_id,path, optional dataset. Run: python convert_h5.py --manifest slides.csv --output means.csv. The converter never overwrites source files.

输出与隐私 / Outputs & privacy

HE评分是冻结LR对C2归属的模型估计,阈值0.5671364068984985,不是个体生存概率。RNA与HE结果独立显示,保留差异;本工具不提供治疗建议。可下载CSV及含版本信息的JSON。页面表格最多显示200例,图最多30例,下载包含全部。

HE score is the frozen LR estimate of C2 membership, not survival probability. Cutoff: 0.5671364068984985. RNA and HE outputs remain independent. Download CSV and versioned JSON; tables show up to 200 cases and plots up to 30, but downloads contain all results.

文件与结果仅存在本次页面内存中;不写入浏览器存储、不上传至计算服务器,不保存分析历史。请使用去标识化编号。清空或关闭页面可移除数据。公开模型参数本身不包含患者级训练数据。

Files and results remain in page memory. No browser-storage persistence, patient-file upload, or analysis history. Use de-identified IDs. Clear or close the page to remove session data. Published model parameters do not contain individual training records.

仅供肺腺癌肿瘤组织科研使用。For lung adenocarcinoma tumor-tissue research only.