單細胞轉錄組確實是利器，但我們大多只利用了表達的信息，而從reads到表達之間的信息完全被我們忽略了。
最近nature發了一篇單細胞方法類文章，講得就是如何利用RNA velocity來做細胞發育路徑的推斷。
velocyto-notebooks
首先需要了解一些基本概念：
RNA velocity：the time derivative of the gene expression state—can be directly estimated by distinguishing between unspliced and spliced mRNAs in common single-cell RNA sequencing protocols. a high-dimensional vector that predicts the future state of individual cells on a timescale of hours. 比較抽象，一開始很難理解。
half-life of mRNA：Translation in both prokaryotes and eukaryotes involves three phases: initiation, elongation, and termination. The relative abundance of nascent (unspliced) and mature (spliced) mRNA can be exploited to estimate the rates of gene splicing and degradation. could reveal the rate and direction of change of the entire transcriptome during dynamic processes.
裝R版本的時候，HDF5老是出問題，必須安裝指定版本。同時還有一系列的預選包必須安裝。
brew switch hdf5 1.10.1_2
conda install llvm
python版本也有問題：
clang: error: unsupported option '-fopenmp'
If not, try conda install llvm, conda install gcc or conda install libgcc.
If this also does not work, try adding -c conda-forge I am pretty sure the version of the compiler provided by conda forge supports openmp.
for pagoda2
cd /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ rm clang++ ln -s /usr/local/clang4/bin/clang++ clang++
合並
ls results/*/*.loom | sed "s:^:`pwd`/: " > loom.file.list
import loompy
files = []
inf = open("loom.file.list", "r")
for line in inf:
files.append(line.strip())
inf.close()
loompy.combine(files, "merged.loom", key="Accession")
import velocyto as vcy
vlm = vcy.VelocytoLoom("merged.loom")
vlm.ca
len(dir(vlm))
待續~
參考：
Messenger RNA Half-Life Measurements in Mammalian Cells