qq_45759229的博客

vscode debug pandas显示多列

ggplot2使用总结

python plotly html转pdf和png

python画图配色总结

ggplot颜色设置总结

我目前python所有的环境基本上都是python3.6.10, 这里安装的scanpy默认是1.7.2，但是安装完，你会发现umap图出现很奇怪的现象，下面是测试的结果%load_ext rpy2.ipython%%R -o counts -o metasuppressMessages(library(splatter))params <- newSplatParams()params <- setParam(params, "nGenes", 5000)params &

R里面的ggplot2画图很好看，但是我对python更熟悉，有没有能再python中画出ggplot2的图呢，是可以的https://towardsdatascience.com/practical-data-visualization-guide-seaborn-vs-ggplot2-9747d9153adepython seabornimport numpy as npimport pandas as pdimport seaborn as snssns.set(style='darkgr

fig, axs = plt.subplots(1, 2, figsize=(12,5),constrained_layout=True)sc.pl.umap(adata, color="tech", title="tech umap", ax=axs[0], show=False)## 注意这个地方一定要show=Falsesc.pl.umap(adata, color="BATCH", title="celltype", ax=axs[1], show=False)## 注意这个地方一定要show=

from sklearn.metrics.cluster import adjusted_rand_score,normalized_mutual_info_scoreimport numpy as np ################ Functions ############## Find optimal resolution with given n_clusters######################################## Attention ##########

USPS数据集可视化import h5pyhf = h5py.File('./data.h5', 'r')X = np.asarray(hf.get('data'), dtype='float32')X_train = (X - np.float32(127.5)) / np.float32(127.5)y_train = np.asarray(hf.get('labels'), dtype='int32')print(X_train.shape)print(y_train)print(np

多数据集pancreas数据集suppressMessages(require(Seurat))suppressMessages(require(ggplot2))suppressMessages(require(cowplot))suppressMessages(require(scater))suppressMessages(require(scran))suppressMessages(require(BiocParallel))suppressMessages(require(Bio

今天测试一个数据时，发现scanpy画的图和使用sklearn画的图有点不一样，解决过程如下测试1from sklearn import datasetsimport scanpy as scimport numpy as npimport randomimport matplotlib.pyplot as pltfrom sklearn.manifold import TSNEnp.random.seed(1)random.seed(1)iris = datasets.load_ir

https://stackoverflow.com/questions/47633546/relationship-between-dpi-and-figure-sizeimport numpy as npimport matplotlib.pyplot as pltx=np.linspace(-2,2)y=np.sin(x)plt.figure(figsize=(10,8),dpi=50)plt.plot(x,y)plt.savefig("./1.png")plt.show()plt.f

直接输入pip install MulticoreTSNE会出现以下错误解决办法：pip install cmake==3.18.4不要安装最新的cmake,否则会出错误最终测试

pythonimport scanpy as scadata=sc.datasets.paul15()sc.pp.normalize_per_cell(adata,counts_per_cell_after=1e4)# 这一部也是需要加入进去的。sc.pp.filter_genes(adata,min_cells=1)sc.pp.filter_genes_dispersion(adata,n_top_genes =2000) #top 1000 gene#log1p datasc.pp.lo

有的时候画图的时候，默认的scanpy画图的大小如下：with rc_context({'figure.figsize':(12,8)}): sc.pl.umap(enc,color=["celltype"]) sc.pl.umap(enc,color=["louvain1.0"],legend_loc="on data")结果如下现在我想把这个数字调大一点方法1with rc_context({'figure.figsize':(12,8)}): sc.pl.umap(

import umapfrom sklearn.datasets import load_digitsimport matplotlib.pyplot as pltX,target= load_digits(return_X_y=True)reducer=umap.UMAP(random_state=0)X_transformed=reducer.fit_transform(X)fig=plt.figure(figsize=(20,12))for label in np.unique(tar

安装conda install -c pyviz holoviews bokeh但是安装完会出现一个问题 File "/home/yxk/anaconda3/lib/python3.8/site-packages/panel/widgets/indicators.py", line 11, in <module> from tqdm.asyncio import tqdm as _tqdmModuleNotFoundError: No module named 'tqdm.a

网上说的ubuntu安装graphviz直接输入sudo apt-get install graphviz但是最后base环境中出现ModuleNotFoundError: No module named 'graphviz'因为我是想在conda的base环境中运行graphviz, 所以正确的安装方式如下：conda install graphvizconda install python-graphviz安装好后，测试如下：from graphviz import Digrap

步骤1插入->形状，获得一个实心圆步骤2用自由曲线画两个封闭区域步骤3全选形状->合并形状->拆分然后再删去不想要的部分，就可以了

看到这篇文章The art of using t-SNE for single-cell transcriptomics，觉得它的图很好看，就看一下文章的结果，其中涉及到安装FIt-SNE，涉及到tsne的加速，下面是FIt_SNE的安装总结我用的是ubuntu20.04，安装的环境是session_info 1.0.0-----Python 3.8.3 (default, Jul 2 2020, 16:21:59) [GCC 7.3.0]Linux-5.11.0-25-gene

library(repr)options(repr.plot.width=10,repr.plot.height=3)print(cowplot::plot_grid(scater::plotTSNE(bct.mBN, colour_by = "clust"), scater::plotTSNE(bct.mBN, colour_by = "cell.class"), scater::plotTSNE(bc