Add GenomicPro multi-omics analysis framework

Project.toml

@@ -0,0 +1,15 @@
+name = "GenomicPro"
+uuid = "0c0bde55-41a4-47af-8c3e-91f915c1c200"
+authors = ["AI Developer <dev@example.com>"]
+version = "0.1.0"
+
+[deps]
+CSV = "336ed68f-0bac-5ca0-87d4-7b16caf5d00b"
+DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
+JSON3 = "0f8b85d8-493c-5b1a-9879-9ecda8c0e6a9"
+
+[compat]
+CSV = "0.10"
+DataFrames = "1"
+JSON3 = "1"
+julia = "1.12"

config/example_pipeline.json

@@ -0,0 +1,42 @@
+{
+  "datasets": [
+    {
+      "name": "transcriptomics",
+      "path": "data/transcriptomics.csv",
+      "delimiter": ",",
+      "feature_column": "Feature",
+      "normalization": "zscore",
+      "imputation": "mean",
+      "metadata": {
+        "missing_strings": ["NA", ""]
+      }
+    },
+    {
+      "name": "proteomics",
+      "path": "data/proteomics.csv",
+      "delimiter": ",",
+      "feature_column": "Feature",
+      "normalization": "zscore",
+      "imputation": "median"
+    }
+  ],
+  "integration": {
+    "strategy": "concatenate"
+  },
+  "analysis": {
+    "run_pca": true,
+    "pca_components": 3,
+    "clustering": "kmeans",
+    "cluster_count": 3
+  },
+  "report": {
+    "output_path": "reports/pipeline_report.json",
+    "text_output_path": "reports/pipeline_report.txt"
+  },
+  "quality_control": {
+    "min_samples": 2,
+    "min_features": 2,
+    "min_nonmissing_ratio": 0.5,
+    "min_variance": 1e-8
+  }
+}

readme.md

@@ -1 +1,79 @@
-meibujun julia
+# GenomicPro
+
+GenomicPro 是一个使用 Julia 语言实现的多组学（multi-omics）分析流水线框架，提供从数据加载、质量控制、归一化、组学数据对齐、集成分析到报告生成的一站式能力。该实现基于 Julia v1.12.1，侧重于易用性与可扩展性，支持 CLI 运行以及通过配置文件管理复杂工作流。
+
+## 关键特性
+
+- **多组学数据加载**：支持从 CSV/TSV 表格读取不同组学层的数据，灵活配置分隔符、特征列、缺失值标记等。
+- **缺失值填补与质量控制**：内置均值/中位数/零值填补策略，支持对低方差、缺失率过高的特征进行过滤。
+- **标准化与变换**：提供 z-score、min-max、robust scaling 以及可选对数变换，适配不同测序平台的特性。
+- **数据对齐与集成**：自动对齐样本，支持特征拼接（concatenate）与加权求和（weighted sum）两种集成策略。
+- **统计/机器学习分析**：内置 PCA 降维与 KMeans 聚类，可根据配置决定是否执行并调整参数。
+- **报告生成**：输出 JSON 与文本摘要，涵盖数据概览、质量指标、分析结果，方便集成到下游系统。
+
+## 项目结构
+
+```
+Project.toml
+src/
+  GenomicPro.jl        # 模块入口
+  types.jl             # 核心类型定义
+  io.jl                # 数据读取与配置解析
+  preprocessing.jl     # 缺失值填补、归一化、对齐等预处理
+  integration.jl       # 多组学集成算法
+  analysis.jl          # PCA、KMeans 等分析
+  reporting.jl         # 报告与摘要生成
+  validation.jl        # 数据校验与质量过滤
+  utils.jl             # 通用辅助函数
+  pipeline.jl          # 流水线编排
+scripts/
+  genomicpro_cli.jl    # 命令行入口
+config/
+  example_pipeline.json# 配置示例
+```
+
+## 安装依赖
+
+```bash
+julia --project -e 'using Pkg; Pkg.instantiate()'
+```
+
+## 运行示例流水线
+
+准备示例数据（可参考 `test/data` 目录中的 CSV），并在根目录创建输出目录：
+
+```bash
+mkdir -p data reports
+cp test/data/*.csv data/
+```
+
+然后运行命令行脚本：
+
+```bash
+julia scripts/genomicpro_cli.jl config/example_pipeline.json
+```
+
+执行完成后，JSON 与文本报告分别保存在 `reports/pipeline_report.json` 与 `reports/pipeline_report.txt`。
+
+## 在代码中使用
+
+```julia
+using GenomicPro
+
+config = read_pipeline_config("config/example_pipeline.json")
+result = run_pipeline(config)
+println(render_summary(result))
+```
+
+## 测试
+
+```bash
+julia --project -e 'using Pkg; Pkg.test()'
+```
+
+## 扩展建议
+
+- 接入更多集成策略（如基于网络的融合、协同矩阵分解等）。
+- 支持批量效应校正与更丰富的归一化方法。
+- 引入更高级的聚类与分类算法，或与 MLJ 生态衔接。
+- 构建交互式可视化报告，支持在浏览器中探索多组学结果。

scripts/genomicpro_cli.jl

@@ -0,0 +1,18 @@
+#!/usr/bin/env julia
+using Pkg
+Pkg.activate(joinpath(@__DIR__, ".."))
+using GenomicPro
+
+function print_usage()
+    println("Usage: genomicpro_cli.jl <pipeline_config.json>")
+end
+
+function main()
+    isempty(ARGS) && return (print_usage(); exit(1))
+    config_path = ARGS[1]
+    config = read_pipeline_config(config_path)
+    result = run_pipeline(config)
+    println(render_summary(result))
+end
+
+main()

src/GenomicPro.jl

@@ -0,0 +1,27 @@
+module GenomicPro
+
+export DatasetConfig, IntegrationConfig, AnalysisConfig, ReportConfig, QualityControlConfig,
+       PipelineConfig, PipelineResult, OmicsDataset, IntegratedDataset,
+       load_omics_dataset, read_pipeline_config, run_pipeline, align_datasets,
+       integrate_datasets, normalize!, impute_missing!, filter_low_quality!,
+       run_pca, run_kmeans, build_report, save_report, render_summary
+
+using CSV
+using DataFrames
+using JSON3
+using LinearAlgebra
+using Statistics
+using Random
+using Dates
+
+include("types.jl")
+include("utils.jl")
+include("validation.jl")
+include("io.jl")
+include("preprocessing.jl")
+include("integration.jl")
+include("analysis.jl")
+include("reporting.jl")
+include("pipeline.jl")
+
+end

src/analysis.jl

@@ -0,0 +1,108 @@
+function run_pca(dataset::IntegratedDataset; components::Int=3)
+    components > 0 || throw(ArgumentError("Number of components must be positive"))
+    n_samples = length(dataset.samples)
+    n_samples > 1 || throw(ArgumentError("PCA requires at least two samples"))
+    mat = dataset.matrix
+    centered = mat .- mean(mat, dims=2)
+    data = transpose(centered)
+    svd_res = svd(data; full=false)
+    k = min(components, length(svd_res.S))
+    scores = svd_res.U[:, 1:k] * Diagonal(svd_res.S[1:k])
+    loadings = svd_res.V[:, 1:k]
+    denom = max(n_samples - 1, 1)
+    explained_variance = (svd_res.S .^ 2) / denom
+    explained_ratio = explained_variance ./ sum(explained_variance)
+    return PCAResult(scores, loadings, explained_variance[1:k], explained_ratio[1:k])
+end
+
+function _initialise_centroids(data::Matrix{Float64}, k::Int, seed::Int)
+    Random.seed!(seed)
+    indices = randperm(size(data, 2))[1:k]
+    return data[:, indices]
+end
+
+function _assign_clusters(data::Matrix{Float64}, centroids::Matrix{Float64})
+    assignments = Vector{Int}(undef, size(data, 2))
+    for j in 1:size(data, 2)
+        column = view(data, :, j)
+        best_idx = 1
+        best_dist = Inf
+        for c in 1:size(centroids, 2)
+            dist = sum((column .- view(centroids, :, c)).^2)
+            if dist < best_dist
+                best_dist = dist
+                best_idx = c
+            end
+        end
+        assignments[j] = best_idx
+    end
+    return assignments
+end
+
+function _update_centroids(data::Matrix{Float64}, assignments::Vector{Int}, k::Int)
+    dim = size(data, 1)
+    centroids = zeros(Float64, dim, k)
+    counts = zeros(Int, k)
+    for j in 1:length(assignments)
+        cluster = assignments[j]
+        centroids[:, cluster] .+= view(data, :, j)
+        counts[cluster] += 1
+    end
+    for c in 1:k
+        if counts[c] > 0
+            centroids[:, c] ./= counts[c]
+        end
+    end
+    return centroids, counts
+end
+
+function _compute_inertia(data::Matrix{Float64}, assignments::Vector{Int}, centroids::Matrix{Float64})
+    total = 0.0
+    for j in 1:size(data, 2)
+        cluster = assignments[j]
+        diff = view(data, :, j) .- view(centroids, :, cluster)
+        total += sum(diff .^ 2)
+    end
+    return total
+end
+
+function run_kmeans(dataset::IntegratedDataset; k::Int=3, maxiter::Int=300, tol::Float64=1e-4, seed::Int=42)
+    k > 0 || throw(ArgumentError("Number of clusters must be positive"))
+    data = dataset.matrix
+    size(data, 2) >= k || throw(ArgumentError("Number of clusters cannot exceed sample count"))
+    centroids = _initialise_centroids(data, k, seed)
+    assignments = _assign_clusters(data, centroids)
+    previous_inertia = Inf
+    for iter in 1:maxiter
+        centroids, counts = _update_centroids(data, assignments, k)
+        for c in 1:k
+            if counts[c] == 0
+                centroids[:, c] = data[:, rand(1:size(data, 2))]
+            end
+        end
+        assignments = _assign_clusters(data, centroids)
+        inertia = _compute_inertia(data, assignments, centroids)
+        if abs(previous_inertia - inertia) < tol
+            previous_inertia = inertia
+            break
+        end
+        previous_inertia = inertia
+    end
+    return ClusteringResult(assignments, centroids, _compute_inertia(data, assignments, centroids))
+end
+
+function run_analysis(dataset::IntegratedDataset, config::AnalysisConfig)
+    pca_result = nothing
+    if config.run_pca
+        pca_result = run_pca(dataset; components=config.pca_components)
+    end
+    clustering_result = nothing
+    if config.clustering === :kmeans
+        clustering_result = run_kmeans(dataset; k=config.cluster_count, seed=config.random_seed)
+    elseif config.clustering === nothing || config.clustering === :none
+        clustering_result = nothing
+    else
+        throw(ArgumentError("Unsupported clustering strategy: $(config.clustering)"))
+    end
+    return AnalysisResult(pca=pca_result, clustering=clustering_result)
+end

src/integration.jl

@@ -0,0 +1,44 @@
+function integrate_concatenate(datasets::Vector{OmicsDataset})
+    combined_matrix = vcat([ds.matrix for ds in datasets]...)
+    combined_features = String[]
+    for ds in datasets
+        append!(combined_features, [string(ds.name, "::", feature) for feature in ds.features])
+    end
+    metadata = Dict{String, Any}(
+        "strategy" => "concatenate",
+        "source_datasets" => [ds.name for ds in datasets]
+    )
+    return IntegratedDataset("Integrated", combined_matrix, combined_features, copy(datasets[1].samples), metadata)
+end
+
+function integrate_weighted_sum(datasets::Vector{OmicsDataset}, weights::Dict{String, Float64})
+    reference_features = datasets[1].features
+    for ds in datasets[2:end]
+        ds.features == reference_features || throw(ArgumentError("Weighted sum integration requires identical feature ordering"))
+    end
+    weight_vec = [weights[ds.name] for ds in datasets]
+    weight_sum = sum(weight_vec)
+    weight_sum == 0 && throw(ArgumentError("Weights must not sum to zero"))
+    normalised = weight_vec ./ weight_sum
+    combined_matrix = zeros(Float64, size(datasets[1].matrix))
+    for (w, ds) in zip(normalised, datasets)
+        combined_matrix .+= w .* ds.matrix
+    end
+    metadata = Dict{String, Any}(
+        "strategy" => "weighted_sum",
+        "weights" => Dict(ds.name => weights[ds.name] for ds in datasets)
+    )
+    return IntegratedDataset("Integrated", combined_matrix, copy(reference_features), copy(datasets[1].samples), metadata)
+end
+
+function integrate_datasets(datasets::Vector{OmicsDataset}, config::IntegrationConfig)
+    isempty(datasets) && throw(ArgumentError("No datasets provided for integration"))
+    if config.strategy == :concatenate
+        return integrate_concatenate(datasets)
+    elseif config.strategy == :weighted_sum
+        weights = _ensure_weights(datasets, config.weights)
+        return integrate_weighted_sum(datasets, weights)
+    else
+        throw(ArgumentError("Unsupported integration strategy: $(config.strategy)"))
+    end
+end