掐架的额外收获

你昨天才做的分析,可能是几年前的结果!》这篇文章给大家敲了警钟,各种各样的web-server,要小心看有没有维护更新,有些是五年十年都不更新的,十分可怕。文章虽然讲的是富集分析,但其它分析工具你同样需要小心。

当然并不是说独立的软件/软件包就一定靠谱,如果软件自己打包了数据,同样要注意数据是否有更新,而如果数据不打包在软件里,而是在线获取,你同样也该留一下心。这也是clusterProfiler做富集分析的优势所在,KEGG数据是在线的,永远是最新的,而GO的数据不在软件包里,而依赖于别的数据包,而这些数据包是社区维护的(相对而言,个人的维护比较难以为继),就确保了数据一直在有更新维护的。

Bioconductor每半年发行一次,注释包同样每半年更新一次,所以你用clusterProfiler做GO分析,你用的GO数据库不会说超过半年没更新,而不像有些公司给出的结果,落后于这个世界不是一年两年这么简单。

Continue reading

听说你想打水印?

正如我在《用R画meme ;)》里说的,meme包就可以拿来画,当然这个包是设计来画meme的,所以文本只会打在中间,如果你想在右下角之类的,不好意思,我没提供选项让你调文本位置,氮素,meme依然是你的好帮手,怎么说呢,meme不单能帮你读图片,还能用grid画出来,这意味着你可以用grid.text随便加个文本就OK了。

实例演示一发:

library(meme)
u <- system.file("angry8.jpg", package="meme")
x <- meme(u)
print(x)
grid.text("biobabble", x=.9, y=.05, 
	gp=gpar(col='white', cex=1.2))

Continue reading

正如我在「知识星球」里说的,scholar这个包现在已经死了,因为作者不维护了,除非出现接盘侠。我自己fork了一个,打了补丁,现在是可用的。这个包应该说我用得蛮多的,你看我简历里的引用数柱状图,我不会无聊到过一段时间就去查,然后重新画图的,太low。都是自动化的,我到简历目录下打make指令,就会自动去爬数据、重新画图并生成简历,而我啥都不用干。

Continue reading

github上的问题,问了两个问题,这是其中第二个:

Meanwhile, when I fed the ENTREZID to enrichKEGG, it show me two unreasonable results:

kegg_enrich = enrichKEGG(
	gene = new_ids$ENTREZID,
	organism = 'hsa',
	keyType = 'ncbi-geneid',
	pvalueCutoff = 0.05,
	#pAdjustMethod = 'BH',
	#qvalueCutoff = 0.2,
	use_internal_data = FALSE
)
head(kegg_enrich)
              ID                           Description GeneRatio  BgRatio       pvalue   p.adjust     qvalue
hsa04380 hsa04380            Osteoclast differentiation    14/281 128/7299 0.0003817040 0.04767338 0.04508266
hsa04070 hsa04070 Phosphatidylinositol signaling system    12/281  99/7299 0.0003875885 0.04767338 0.04508266

I noted that only 281 genes are remained(there are 700+ genes in my list). In case that there is something wrong wtihin my gene list, I also tried my list with DAVID. It gave me reasonable results. So this is my second question, why enrichKEGG cannot recognize my geneids?

Continue reading

在很久很久以前,当我还在用perl的时候,有一个TGen-EUtils的perl模块,可以去爬PubMed,我喜欢用它来画所谓的pubmed趋势,比如说我在2012年的时候,就画了各个组学技术的pubmed趋势,通过每一年的文章发表量来看某个技术到底火不火,走势怎么样。

当年microarray还特别猛,但显然可以看到二代测序必须要火,也可以看到proteomics在吹了几年牛B之后,有回落的趋势。

Continue reading

After the release of meme package, I received several feedbacks from users.

The most usefule one is the comment on my blog post:

Greetings Mr. Yu,

I am very happy that this package exists. Thank you for making it! I would like to request a feature, to ensure the package is able to compete with professional meme-creation tools like memegenerator and paint.net. Since memes often use the font Impact, in white and with black outline, I believe the package would be more powerful if it also did that automatically.

Regards,

The words, ‘compete with professional meme-creation tools’, stimulated me to develop text plotting with background outline effect.

Now this feature is available in meme v>=0.0.7, which can be downloaded from CRAN.

Continue reading

Sercan Kahveci

Greetings Mr. Yu,

I am very happy that this package exists. Thank you for making it! I would like to request a feature, to ensure the package is able to compete with professional meme-creation tools like memegenerator and paint.net. Since memes often use the font Impact, in white and with black outline, I believe the package would be more powerful if it also did that automatically.

Regards,

Sercan Kahveci, MSc

Content creator at Questionable Research Memes on Facebook https://www.facebook.com/QResearchMemes/

上次一篇《用R画meme ;)》把脸书在做MEME的人都给炸出来了,专业人士给出专业建议,要用Impact字体,并且要让字体出现阴影效果,这真是厉害了,感觉有点神奇,但他一句,这样就可以和专业工具PK了,着实太刺激了,我决定试一试能不能搞个阴影效果。结果还真让我搞成了,文本有一种浮在上面的错觉。

我们知道meme包是grid写的,我的输出是grob对象,当我半夜调试到2点,带着兴奋入睡之后,一点都睡不安稳,早上早早就迷迷糊糊在床上想着,既然我造出了文本阴影效果的grob对象,我应该封装成geom图层,送给ggplot2用户。

Continue reading

Author's picture

Guangchuang Yu

Bioinformatics Professor @ SMU

Bioinformatics Professor

Guangzhou