把研究问题说清楚,
分析交给 Ragomics
Describe the question.
Ragomics runs the analysis.
不用写代码,也不用自己搭流程。用自然语言描述研究目标,平台先和你把实验设计、数据来源与结论边界确认清楚,再自动编排并执行分析,交付出版级图表、综合报告和可复现脚本。 No code and no pipeline to build. Describe your research goal in plain language. The platform first confirms the design, the data source and the limits of the conclusions with you. Then it plans and runs the analysis, and delivers publication-grade figures, a full report and reproducible scripts.
data/ qc_summary.csv · spatial_domains.csv …
scripts/ 01_qc … 06_svg · README
像和一位生信同事对话,而不是填表跑流程Like working with a bioinformatician, not filling in a pipeline form
每个关键节点都由你拍板。平台给候选,你选择确认,确认后才进入下一步;运行中随时可以停止。You decide at every key point. The platform proposes, you confirm, and only then does it move on. You can stop a run at any time.
描述目标Describe
一句话或一段话说明想研究什么,可附上自己的数据。State the goal in a sentence or a paragraph. Attach your data if you have it.
澄清设计Clarify
确认疾病与组织、数据来源、对照与重复;样本量不足时直说只能做描述性结论。Confirm tissue, data source, controls and replicates. If the sample size is too small, it says so.
选择确认Confirm
方案、数据集、运行配置以编号选项呈现,你回复序号或说明即可。Plans, datasets and run settings come as numbered options. Reply with a number or a note.
自动执行Run
平台编排整条流程并按任务弹性调度算力,实时显示状态与耗时。The platform runs the full workflow on elastic compute and shows live status and elapsed time.
交付前自检Inspect
逐个核对报告数字、图文一致与交付完整性,发现问题先修正。Every number, figure and file is checked before delivery. Problems are fixed first.
交付成果Deliver
图表、数据表、综合报告与分步脚本,一个压缩包带走。Figures, tables, the report and step-by-step scripts in one package.
完整的生信分析全流程,不限组学End-to-end bioinformatics, across omics
从数据获取、质控到下游解读和成文,覆盖常见组学方向与多组学整合。From data retrieval and QC to downstream interpretation and writing, across the common omics types and their integration.
单细胞Single-cellscRNA
质控、聚类、注释、细胞通讯、轨迹、整合去批次QC, clustering, annotation, cell–cell communication, trajectories, integration
空间转录组SpatialST
空间域、去卷积、空间可变基因、空间通讯、图像分析Spatial domains, deconvolution, spatially variable genes, spatial communication, imaging
bulk 转录组Bulk RNA-seqRNA
差异表达、GO / KEGG / GSEA 富集、时序分析Differential expression, GO / KEGG / GSEA, time series
基因组变异Genomic variantsDNA
变异检测与注释、拷贝数、临床解读Variant calling and annotation, copy number, clinical interpretation
表观组EpigenomicsEPI
甲基化、ATAC-seq、ChIP-seqMethylation, ATAC-seq, ChIP-seq
宏基因组 / 微生物组MetagenomicsMETA
物种与功能注释、多样性、差异丰度Taxonomic and functional profiling, diversity, differential abundance
蛋白组 / 代谢组Proteomics / MetabolomicsMS
定量、差异、通路Quantification, differential analysis, pathways
免疫组库Immune repertoireTCR/BCR
克隆型、多样性、克隆扩增Clonotypes, diversity, clonal expansion
多组学整合与写作辅助Integration and writingMULTI
多组学联合分析;Nature 风格制图、论文润色、引用管理、论文转 PPTJoint multi-omics analysis; Nature-style figures, editing, citations, paper to slides
平台已累计完成上千次分析服务,常见的分析需求都能覆盖。The platform has delivered more than a thousand analysis services across a wide range of requests.
没有数据?平台帮你找No data? The platform finds it
不用自己翻数据库、下载文件。No need to search databases or download files yourself.
- 系统检索公开数据库,按研究问题筛选候选Searches public databases and shortlists candidates for your question
- 识别测序平台、物种与实验设计(如 Visium、CosMx、Xenium、GeoMx、MERFISH)Identifies platform, species and design (Visium, CosMx, Xenium, GeoMx, MERFISH …)
- 核实数据是否公开、是否受控访问,再交给你确认Checks whether data are public or controlled-access before asking you
自有数据,多种方式接入Your own data, many ways in
上传后直接在对话里使用。Upload and use it in the conversation.
- 本地文件、整个文件夹上传,拖拽或粘贴即可Upload files or whole folders, by drag and drop or paste
- 百度网盘、夸克网盘直接导入Import directly from Baidu Netdisk and Quark
- 复用历史会话的文件和分析输出,关联其他会话Reuse files and outputs from earlier sessions, or link sessions
交付之前,先自己查一遍Checked before it reaches you
「智能自检」在交付前对成果做多维度核验。发现问题会原路修正、重建受影响的产物并复核,不改动源数据。Smart inspection checks the results on several independent dimensions before delivery. Problems are fixed at the source, affected outputs are rebuilt and re-checked, and source data are never changed.
数字可溯源Traceable numbers
报告里每个关键数字都从结果文件独立重算、逐一比对。Every key number is recomputed from result files and compared.
图文一致Figures match text
逐张图核对与正文结论、图注是否一致,并检查图面。Each figure is checked against the text, its caption and its layout.
表述准确Accurate wording
核对报告与结果表是否完整对应、措辞是否准确。The report is checked against the result tables for coverage and wording.
交付完整Complete package
报告、数据、脚本、索引齐备且相互对应。Report, data, scripts and index are all present and consistent.
逐页复核Page-by-page review
对最终 PDF 报告逐页做视觉复核。The final PDF report is reviewed visually, page by page.
设计阶段把关At design time
动手前确认设计可行性和结论边界,报告中写明能支撑和不能支撑的结论。Feasibility and the limits of the conclusions are agreed before any work, and stated in the report.
执行阶段透明At run time
工作区保留全部过程与中间文件,按目录可浏览、可下载、可追溯。The workspace keeps every intermediate file, browsable and downloadable.
交付阶段核验At delivery
多维自检加可复现脚本,发现问题闭环修正后再交付。Multi-dimension checks plus reproducible scripts. Issues are closed before delivery.
失败的结果也算数Negative results count too
一项基于公开帕金森病与阿尔茨海默病数据的研究项目:分析计划和判定规则在运行前固定,所有结果——包括不支持、无法判定和缺数据的——全部公开。A research campaign on public Parkinson's and Alzheimer's data. Plans and decision rules were fixed before each run, and every outcome is published, including null, inconclusive and untestable ones.
从一个开放问题出发,提名透明细胞肾癌的耐药空间生态位Nominating resistance niches in clear cell renal cell carcinoma from one open question
没有预设机制或基因。整合单细胞(n=19)、空间转录组(n=24)与 TCGA-KIRC(n=613)数据,提出候选耐药生态位与联合用药假设;患者活检的多重免疫荧光在蛋白水平检验了所提组织结构的合理性。No mechanism or gene was given. Single-cell (n=19), spatial (n=24) and TCGA-KIRC (n=613) data were combined to nominate resistance niches and combination hypotheses. Multiplex immunofluorescence on patient biopsies checked the proposed tissue structure at protein level.
同一个纤维化问题,一句话提问与专家级提问的差别The same fibrosis question, asked in one line and as an expert brief
两种提问都找到了跨器官保守的成纤维细胞基质程序,但在巨噬细胞程序和三细胞生态位上结论不同。更严格的专家标准推翻了较宽松的结论。Both found conserved fibroblast matrix programmes across organs, but disagreed on the macrophage programme and the three-cell niche. The stricter expert criteria overturned the looser claim.
三轮专家反馈,与参考研究的一致度逐轮提升Three rounds of expert feedback raise agreement with a reference study
在 IDH 突变型星形细胞瘤公开数据(35 例、五层组学)上,对照参考研究的 17 项结论评估。系统不变,只增加专家输入;提升反映的是吸收反馈的能力。On public IDH-mutant astrocytoma data (35 samples, five omics layers), results were scored against 17 findings of a reference study. The system stayed the same; only expert input grew.
以上结果均为假设生成性结论,联合用药为待检验假设而非建议。数据与报告详见 All results are hypothesis-generating; drug combinations are hypotheses for testing, not recommendations. Data and reports: omicairn.github.io
拿到的是能复现的成果,不只是几张图You get reproducible work, not just figures
├─ report/ # 综合报告 Markdown + PDF,图表内嵌# full report, Markdown + PDF
├─ data/ # 结果数据表# result tables
├─ scripts/ # 分步脚本 + 环境依赖# step scripts + environment
└─ README.md # 目录、图表索引、复现说明# index and how to reproduce
- 结果按「中间 / 最终」分层、按分析步骤分组,可搜索、筛选、在线预览、批量下载Results are split into interim and final, grouped by step, searchable and previewable
- 报告含方法、结果与关键发现,并单独写明本次分析的适用范围The report covers methods, results and key findings, and states the scope of the analysis
按需选择分析档位与算力Choose the tier and the compute
轻量任务与重计算任务可用不同规格,平台按任务弹性调度,无需自建环境。Light and heavy steps can use different specs. Compute is scheduled per task; no setup needed.
把下一个分析交给 RagomicsHand your next analysis to Ragomics
论文支撑、基金申请、临床研究、研发项目——从一句话开始。Papers, grant applications, clinical studies, R&D — start with one sentence.