Laplacian和PCA貌似是同一种性质的方法,坐标系变换。只是拉普拉斯属于图论的范畴,术语更加专业了。
要看就把一篇文章看完整,再看其中有什么值得借鉴的,总结归纳理解后的东西才是属于你的。
问题:
1. 这篇文章有哪些亮点决定他能发NM?单细胞,consensus,较好的表现,包装了一些专业的术语,显得自己很专业,其实真正做的东西很少;
2. consensus方法的本质是什么?
3. 工具的评估准则?ARI,silhouette index
4. SC3的最大缺点是什么?速度太慢,超过1000个细胞就非常耗费计算和存储资源
5. 能看懂SC3这个R包的逻辑吗?核心的就4步,多种距离度量,转换,kmeans聚类,consensus;
The main sc3 method explained above is a wrapper that calls several other SC3 methods in the following order:
- sc3_prepare
- sc3_estimate_k - Tracy-Widom theory - random matrix theory (RMT)
- sc3_calc_dists
- sc3_calc_transfs
- sc3_kmeans
- sc3_calc_consens
- sc3_calc_biology
6. 有很多问题没有回答,这篇文章偏技工!核心就是kmeans,打了个复杂的包而已。
- 不同距离的度量有什么差异?
- 为什么要做两种转换PCA和laplacian?
- 为什么选择了kmeans?不知道它有天然的劣势吗
- 做consensus的理论依据是什么?凭什么说做了一致性后结果就更好?
最近在看SC3聚类这篇文章,SC3使用了这个工具。
SC3: consensus clustering of single-cell RNA-seq data
All distance matrices are then transformed using either principal component analysis (PCA) or by calculating the eigenvectors of the associated graph Laplacian (L = I – D–1/2AD–1/2, where I is the identity matrix, A is a similarity matrix (A = e–A′/max(A′)), where A′ is a distance matrix) and D is the degree matrix of A, a diagonal matrix that contains the row-sums of A on the diagonal (Dii = ΣjAij). The columns of the resulting matrices are then sorted in ascending order by their corresponding eigenvalues.
先看下该工具的功能:SC3 package manual
跑一下常规代码: