Paper sharing
scBatch: Batch Effect Correction of RNA-seq Data through Sample Distance Matrix Adjustment
papar info: 2020; published in Bioinformatics
Novelty
- unfinish
- not seq error correction, it is the correction for gene expression counts
Brief introduction
1.Batch effect: batch means a number of the factors (including different laboratories, different sample preparation batches, different sequencing…)
2.Existing Methods:
- establish linear models of gene expression with biolofical groups (e.g. disease and control group)
- DE gene analysis package (Limma).
2017- scPLS
- DE gene analysis package (Limma).
- empirical Bayes algorithm (e.g. ComBat)
- an improved version 2020-ComBat-seq (by nagative binomial regression)
- consensus clustering method (SC3: conduct clustering analysis based on multiple distance metrics)
- BatchEffectRemovel.
3.Limitation of linear methods: require the knowledge of biological groups for each observation, which is hardly feasible.
4.spike-in genes?
spike-in方法，指在RNA-seq建库过程中参入一些预先知道序列信息以及序列绝对数量的内参，这样在进行RNA-Seq测序的时候就可以通过不同样本之间内参（spike-in）的量来做一条标准曲线，就可以非常准确地对不同样本之间的表达量进行矫正。
常用的spike-in类型：ERCC Control RNA （External RNA Controls Consortium）.
5.workflow
- input： gene expression counts matrix
6.Dataset
- simulated data : (PROPER for data simulation)
- real data
