Self-contained, run-out-of-the-box R & Python modules for everyday bioinformatics. Each module ships a tiny example dataset, runs from a single command, and renders journal-style vector figures. Copy any one of them and its
READMEtells you exactly what data to feed in, what analysis it does, and what figures come out.开箱即跑的 R / Python 生信分析模块库。 每个模块自带小型示例数据,一条命令即可运行, 输出顶刊风格矢量图。复制走任意一个,看它的
README就知道:喂什么数据、做什么分析、出什么图。
What you get · How a module works · Quick start · Status legend · ★ New in 2026 · Example figures · All categories · Framework · Reproducibility · License
EN — A practical toolbox for dry-lab bioinformatics: ~145 modules across 23 analysis categories (single-cell, spatial, Mendelian randomization, WGCNA, ML signatures, docking/MD, enrichment, and more). Most run turnkey on bundled or synthetic example data with no edits; the rest keep their real script + dependency notes for a server/GPU. A shared plotting framework gives every figure the same journal-ready look, and a static linter + quality checklist keep pipelines reproducible.
中文 — 一套实用的干实验生信工具箱:横跨 23 个分析类别(单细胞、空间转录组、孟德尔随机化、 WGCNA 共表达、机器学习签名、对接/分子动力学、富集等)共约 145 个模块。大多数开箱即跑 (自带或合成的示例数据,无需改动);少数需要服务器/GPU 的也保留了真实脚本与依赖说明。 共享绘图框架让所有图保持统一的顶刊审美,静态检查器 + 质量清单保证管道可复现。
Design principles · 设计原则
- Reuse, never from scratch · 复用而非从头写 — start from a real module or a real published tool; never hand-write analysis from memory (that invites hallucinated APIs). 从既有模块或真实已发表工具起手,绝不凭记忆手写分析(那会产生臆造的 API)。
- Honest baselines built in · 内置诚实基线 — every deep-learning / complex model module ships a simple baseline (PCA / linear / LASSO / permutation null), because in 2026 these often don't beat the baseline — the module reports both so you can judge. 每个深度学习/复杂模型模块都配一个简单基线(PCA / 线性 / LASSO / 置换 null);2026 年这些 方法常常打不过基线,模块会把两者都报出来让你判断。
- No plain bar charts · 不用平凡条形图 — top journals rarely use them; modules default to lollipop / dot / dumbbell / violin / raincloud / heatmap / network. 顶刊很少用条形图,模块默认改用棒棒糖图 / 点图 / 哑铃图 / 小提琴 / 云雨图 / 热图 / 网络图。
- Reproducible · 可复现 — fixed seeds, relative paths only, vector PDF + 300 dpi PNG, environment snapshots, CI lint gate. 固定随机种子、仅用相对路径、矢量 PDF + 300dpi PNG、 环境快照、CI 检查门。
Every module is a self-contained folder. 每个模块都是一个自带一切的文件夹:
<category>/<NNN_module>/
├── <NNN_module>.R | .py # main script · 主脚本 (runs on example_data/ by default)
├── README.md # input spec, method, outputs · 输入规格 / 方法 / 输出图
├── example_data/ # small synthetic input · 小型合成输入(可直接跑通)
└── assets/ # committed preview figures · 提交进库的预览图
EN — Run it as-is to render the bundled example; point --input / --outdir at your
own data to reuse it. Open the module's README.md to see the exact input format, the
method (and its honest baseline), and every output figure with a preview. Run-time
outputs (results/) are git-ignored.
中文 — 直接运行就出示例图;用 --input / --outdir 指向你自己的数据即可复用。打开
模块的 README.md 就能看到:确切的输入格式、方法(及其诚实基线)、每一张输出图的预览。
运行时产物(results/)默认被 git 忽略。
git clone https://github.com/fsy2004/bioinfo-reusable-code.git
cd bioinfo-reusable-code/modules
# run a module on its bundled example data · 用自带示例数据运行
Rscript 03_transcriptomics_deg/010_geo_deg_volcano_heatmap_pca/010_*.R
python 08_singlecell_spatial_trajectory/543_squidpy_spatial_statistics/543_*.py
# run on your own data · 换成你自己的数据
Rscript 03_transcriptomics_deg/010_geo_deg_volcano_heatmap_pca/010_*.R \
--input your_matrix.csv --outdir results/run1Tested with R 4.4 and Python 3.12. 已在 R 4.4 / Python 3.12 下测试。
Per-module dependencies are listed in each module's README.md.
每个模块的依赖见其 README.md。
| Mark | Meaning · 含义 |
|---|---|
| ✅ | Turnkey — runs locally on bundled/synthetic data, no edits · 开箱即跑,无需改动 |
| 🟡 | Honest baseline / core runs locally; full method needs the package on a server · 本地跑诚实基线/核心,完整方法需服务器装包 |
| 🔴 | Heavy / GPU / external toolchain (GROMACS, deep-learning FMs) — reference wrapper · 重型/GPU/外部工具链,参考封装 |
| 📄 | Template — bring your own data + install · 模板,自备数据与安装 |
| 📦 | Vendored third-party package · 内置的第三方包源码 |
Full per-module index (purpose, input→output, deps, figure types) + a figure-type → module
reverse index live in modules/CATALOG.md.
逐模块完整索引(用途、输入→输出、依赖、图类型)与**「图类型→模块」反查表**见
modules/CATALOG.md。
EN — 30 modules wrapping 2026-H1 new methods across seven analysis lines. Each is
grounded in the real published package, run-verified on synthetic data, and ships an
honest simple-baseline comparison (FMs/complex models often don't beat PCA/linear/LASSO
— the module reports the baseline). Format: NNN package — what it does (its niche) ·
output figures · status.
中文 — 30 个模块封装 2026 年上半年的新方法,覆盖七条分析线。每个都接地于真实已发表的
软件包、在合成数据上跑通验证,并内置诚实基线对照(基础模型/复杂模型常打不过 PCA/线性/LASSO,
模块会把基线一并报出)。格式:编号 包名 — 做什么 (工具侧重点) · 输出图 · 状态。
557sccomp— Bayesian beta-binomial differential composition test （侧重组成性、防比例假阳） · boxplot+lines · lollipop · raincloud · ✅559muscat— multi-sample pseudobulk differential state （样本级聚合,2026 金标准,防伪重复) · MDS · pathway-volcano · heatmap · lollipop · ✅558miloR— KNN-neighbourhood differential abundance, no clusters （连续状态、邻域级 DA) · beeswarm · network · volcano · 🟡560copyKAT— scRNA copy-number / aneuploidy calling （inferCNV 已弃用后的替代) · CNV heatmap · UMAP · lollipop · ✅532SCpubr— one-call publication-grade single-cell figures （统一顶刊审美、色盲友好) · UMAP · dot · alluvial · ✅
541BANKSY— spatial domain segmentation （自表达+邻域+方位梯度,非DL可解释) · domain map · ARI vs baseline · UMAP · ✅542nnSVG— spatially variable genes (nearest-neighbour GP) （区分空间结构 vs 单纯高变) · spatial expr · lollipop · scatter · ✅543squidpy— spatial statistics: Moran / neighbourhood / Ripley （带置换 null 基线) · nhood heatmap · Moran lollipop · co-occurrence · ✅544PASTE— multi-slice alignment by optimal transport （切片配准、3D 堆叠) · before/after scatter · overlay · ✅545SPOTlight— spot deconvolution + scatterpie （带已知比例 RMSE 校验) · spatial scatterpie · heatmap · scatter · ✅531LIANA+— consensus cell-cell communication （统一多方法、共识更稳) · L-R dotplot · chord · tile · ✅
534MendelianRandomization— MVMR-cML constrained-ML multivariable MR （抗相关+非相关多效性) · forest · dumbbell · heatmap · ✅535MRBEE— bias-correcting cis / MVMR estimator （校正测量误差偏倚) · lollipop · forest · scatter · ✅533MRcare— winner's-curse-free robust MR （内生化选择偏倚) · lollipop · forest · scatter · 🟡536MR-link-2— single-region cis-MR, pleiotropy-robust （单关联区域、控假阳) · forest · LD heatmap · scatter · 🟡537SharePro— effect-group colocalization （多因果变异共定位) · locuscompare · dot · heatmap · 📦
538NetRep— cross-dataset module preservation permutation test （模块可重复性、外部验证) · Zsummary scatter · lollipop · null density · ✅539SmCCNet— phenotype-driven multi-omics sparse-CCA network （性状特异跨组学子网) · network · adjacency heatmap · lollipop · ✅540CWGCNA— causal-direction module↔trait mediation （回应「共表达=相关≠因果」) · forest · network · lollipop · 🟡
554RobustRankAggreg— consensus feature selection by robust rank aggregation （稳定性>单方法) · lollipop · rank heatmap · UpSet · ✅550TabPFN— tabular foundation model classifier （小样本能打过 GBDT 的 2026 硬证据;折内防泄漏) · ROC+PR · calibration · lollipop · ✅551aorsf— accelerated oblique random survival forest （比标准 RSF 更准、对照 Cox) · time-AUC · importance lollipop · KM · ✅552survex/SurvSHAP(t)— time-dependent survival explanation （特征贡献随时间变化) · SurvSHAP curve · importance · BD · ✅553riskRegression— calibration + decision curve + time-AUC （Stop-Chasing-C-index:补校准与临床获益) · calibration · DCA · time-AUC · ✅555MAPIE/crepes— conformal prediction UQ （给签名预测统计有效的覆盖保证) · calibration scatter · set-size violin · coverage · dumbbell · ✅
547ProLIF— protein-ligand interaction fingerprint （逐帧占据率、客观非主观挑残基) · barcode · interaction heatmap · lollipop · ✅548bio3d— MD DCCM / PCA / RMSF ensemble analysis （集体运动与柔性区) · DCCM heatmap · PCA scatter · RMSF lollipop · ✅556PoseBusters— docking-pose physical validity gate （>50% DL pose 物理无效→守门) · tick heatmap · per-check lollipop · dumbbell · ✅
546enrichplot— dotplot / cnetplot / emapplot / treeplot （代替富集条形图;cnetplot 已迁 ggtangle) · dot · gene-concept network · module map · tree · ✅549GOplot— chord / circle enrichment figures （基因-通路多对多关系) · GOChord · GOCircle · GOHeat · ✅
Background, dated sources, and "does it beat PCA/linear?" notes for these 2026 tools are compiled in the companion knowledge base (
bioinfo-DL-library/analysis-tools-2026/). 这些 2026 工具的背景、发表年月与「打不打得过基线」的提醒,汇编在配套知识库bioinfo-DL-library/analysis-tools-2026/。
Rendered directly from bundled/synthetic example data · 直接由自带/合成示例数据渲染:
| DE volcano · 差异火山 | scRNA UMAP · 单细胞 | MR scatter · 孟德尔随机化 |
|---|---|---|
 |
 |
 |
| Spatial domains · 空间域 (BANKSY) | Composition · 组成 (sccomp) | Enrichment network · 富集网络 (cnetplot) |
|---|---|---|
 |
 |
 |
| Spatial niche · 空间邻域 (squidpy) | Diagnostic ROC · 诊断 (TabPFN) | Decision curve · 决策曲线 (riskRegression) |
|---|---|---|
 |
 |
 |
| # | Category · 类别 | Typical output · 典型输出 |
|---|---|---|
| 01 | Network pharmacology & target DBs · 网络药理与靶点库 | Venn, UpSet, target tables |
| 02 | GO / KEGG enrichment · 富集 | dot, cnetplot, emapplot, chord |
| 03 | Transcriptomics & DE · 转录组与差异表达 | volcano, heatmap, PCA, pseudobulk DS |
| 04 | ML feature selection · 机器学习特征选择 | LASSO, RF, SHAP, consensus, UpSet |
| 05 | Diagnostic models · 诊断模型 | ROC, calibration, DCA, nomogram, TabPFN |
| 06 | Immune infiltration · 免疫浸润 | composition, boxplot, deconvolution |
| 07 | Docking & MD · 对接与分子动力学 | interaction fingerprint, DCCM, PoseBusters |
| 08 | Single-cell / spatial / trajectory · 单细胞/空间/轨迹 | UMAP, dotplot, BANKSY domains, sccomp, copyKAT |
| 09 | Mendelian randomization & GWAS · 孟德尔随机化 | scatter, forest, MVMR-cML, MRBEE, coloc |
| 10 | TWAS (single-cell eQTL) · 单细胞 eQTL 权重 | weight tables |
| 11 | WGCNA co-expression · 共表达网络 | module-trait heatmap, NetRep, SmCCNet |
| 12 | TCGA prognosis · 预后 | KM, time-ROC, aorsf, SurvSHAP, DCA |
| 13 | TF regulation / circos · 转录因子调控 | chromosome circos, regulon network |
| 14 | Single-cell in-silico perturbation · 虚拟扰动 | gene-knockout effects, Geneformer |
| 15 | Drug perturbation / repurposing · 药物扰动 | pharmacovigilance, beyondcell |
| 16 | Spatial communication / fate · 空间通讯与命运 | CellRank, niche, LIANA+, PASTE, SPOTlight |
| 17 | Advanced result figures · 高级结果图 | raincloud, ridgeline, dumbbell, chord |
| 18 | External method sources · 外部方法源 | manifest only |
| 19 | Multi-omics integration · 多组学整合 | MOFA, consensus clustering |
| 20 | Mutation / CNV / methylation / proteome · 变异/甲基化/蛋白组 | oncoprint, volcano, heatmap |
| 21 | Disease burden (GBD / NHANES / CHARLS) · 疾病负担 | ASR/EAPC, survey-weighted, comorbidity network |
| 22 | Single-cell metabolism · 单细胞代谢 | metabolic pathway activity |
| 23 | Uncertainty & conformal prediction · 不确定性与共形预测 | coverage, calibration, prediction-set size |
Categories 10, 14, parts of 07/16 need heavy/GPU toolchains (FUSION, GROMACS, DL FMs);
their scripts + dependency notes are kept for reference. Modules marked 🟡 run a real
baseline/core locally and need the full package on a server — see
modules/_framework/SERVER_DEPENDENCIES.md.
类别 10、14 及 07/16 的部分需要重型/GPU 工具链;🟡 模块本地跑真实基线/核心,完整方法需服务器装包,
详见 SERVER_DEPENDENCIES.md。
Shared by all modules so figures and I/O stay consistent · 所有模块共用,保证图与 I/O 一致:
theme_pub.R/pubstyle.py— Nature-aligned theme; journal palettes (NPG / AAAS / Lancet + colour-blind-safe Okabe-Ito), viridis for continuous, RdBu for diverging;save_fig()exports vector PDF + 300 dpi PNG. 顶刊主题与配色,一次导出矢量 PDF + 300dpi PNG。CONVENTIONS.md— module layout, turnkey rules, figure rules · 模块结构、开箱即跑、绘图规范。ANALYSIS_TEMPLATE/— scaffold for a new multi-step project (central config, seed, checkpointed steps, env snapshot; R + Python) · 新项目脚手架。QUALITY_CHECKLIST.md— pre/in/post-analysis checklist · 分析前/中/后质量清单。qc_lint.py— static checks for hard-coded paths, missing seeds, non-vector exports, missing env snapshots; also a CI gate · 静态检查 + CI 门。TOOL_SELECTION_GUIDE.md— pick the right module/tool for a task · 任务→模块/工具选择指南。
- Modules run on bundled example data with no edits; use
--input/--outdirto switch. 模块用自带示例数据零改动即跑;用--input/--outdir切换。 - No absolute paths or
setwd(); figures exported as vector PDF + 300 dpi PNG. 不用绝对路径或setwd();图导出为矢量 PDF + 300dpi PNG。 - Fixed seeds; honest baselines reported alongside complex models; figure text in English, code comments bilingual. 固定随机种子;复杂模型旁报诚实基线;图中文字英文、代码注释中文。
- Reuse the framework instead of re-implementing themes or I/O. 复用框架而非重写主题与 I/O。
Each module follows the license of the tools and methods it uses. Vendored third-party code
keeps its original license — see the relevant module README and upstream repository.
每个模块遵循其所用工具与方法的许可。内置的第三方代码保留其原始许可,详见对应模块 README
与上游仓库。