由买买提看人间百态

topics

全部话题 - 话题: replicate
首页 上页 1 2 3 4 5 6 7 8 9 10 下页 末页 (共10页)
p*****2
发帖数: 21240
1
怒赞。貌似是主流startup的标配阿。
你们是拿Redis当cache用,MongoDB存transactional data?
我们的数据分三种,一种需要非常快速的访问,就用Redis, 一种需要很灵活的query,
就用Mongo。现在数据量还不够大,大了以后会考虑上Hbase了。我们没有transaction
。或者说schema设计的适合nosql,不需要transaction了。
另外,据说,node JS的performance不是很好,支持不了访问量很大的网站?
我们不是网站,是service。以前是Ruby的,现在是node。效果比以前好很多。你说支
持不了访问量很多大的网站什么意思?我们的主网站就是node的,没发现有什么问题,
rail都可以,node会不行吗?
然后Redis和MongoDB在cross DC replication上做的不太好?是不是真的啊?
我们redis replication没有cross DC。两个DC个有一个instance。本来这东西就很占
内存,很多时候需要partition。
mongo replicatoin across D... 阅读全帖
v*****r
发帖数: 1119
2
Ceph 的冗余控制好像没有用 RAID 5 的机制, 我的理解是靠维护multiple copies
across the distributed storage, 我猜设置里会让你 configure how many copies
you want to keep for 冗余控制, 一个 node 挂了,replication process then
kick off to restore the damaged 冗余控制 copy. To speed up the replication
across the network, RAID 0 肯定是用了。
c***d
发帖数: 996
3
来自主题: Programming版 - 今天碰见一个高手
先看个video:
http://www.youtube.com/watch?v=hEqQMLSXQlY
他主持的一个系统,replication策略我觉得和hdfs有重复, 我就问为什么不直接架在
hdfs上作。这位大哥大概是这么解释的:
hdfs的metadata management其实是single point of failure,确实可以用backup
namenode来改善,但这不是问题的根本。问题的根本是distributed storage的
replication information不应该用lookup 来解决,而应该用比较robust的hash
function。那会不会不elastic呢? 确实有这个问题,这个问题的关键在于,系统本身
要有一个作elastic logic的controller, 随时得到整个系统的信息。 当有node增加进
来或者fail掉,elastic logic controller会根据这个当前情况计算出理想的
distribution, 并开始移动block。移动block是copy and delete, 系统在近opti... 阅读全帖
c***d
发帖数: 996
4
来自主题: Programming版 - 今天碰见一个高手
回去看aws 65648 report, 把我给看乐了:
"Two factors caused the situation in this EBS cluster to degrade further
during the early part of the event.
... ...
There was also a race condition in the code on the EBS nodes that, with a
very low probability, caused them to fail when they were concurrently
closing a large number of requests for replication. In a normally operating
EBS cluster, this issue would result in very few, if any, node crashes;
however, during this re-mirroring storm, the volume of connect... 阅读全帖
g*****g
发帖数: 34805
5
来自主题: Programming版 - C# is light-years ahead of Java now
You are wrong. And you've been wrong every time. But this?
You thought Nasdaq was running on 2 4nodes MS SQL server and that's deadly
wrong. And you have no experience on scalability whatsoever and you are
talking like scaling something to 1m transactions/s is as simple as 5000/s
as long as you have 200 more times hardware there. You don't know how the
hardware compares. You don't know the latency requirement, you don't know
the redundancy requirement. But you compare nonetheless just because yo... 阅读全帖
f*******t
发帖数: 7549
6
来自主题: Programming版 - 坛子里有人搞HBase的吗?
我们有三种replication:
1.异步replay log的slave
2.client同时写两个hbase,只从其中一个读取,对data loss和availability要求不高
3.quorum multi data center synchronous replication,快要发布了

★ 发自iPhone App: ChineseWeb 8.6
f*******t
发帖数: 7549
7
来自主题: Programming版 - 坛子里有人搞HBase的吗?
我们有三种replication:
1.异步replay log的slave
2.client同时写两个hbase,只从其中一个读取,对data loss和availability要求不高
3.quorum multi data center synchronous replication,快要发布了

★ 发自iPhone App: ChineseWeb 8.6
i**i
发帖数: 1500
8
http://www.infoworld.com/print/236070
Meet AOSP, the other Android, while you still can
By Galen M. Gruman
Created 2014-02-14 04:00AM
No doubt you've seen the stat showing how much more popular Android
smartphones are than iPhones in much of the world. But many of those Android
smartphones aren't the Android you're thinking of -- the kind you'd get
from Samsung, HTC, or Motorola. That's because there's more than one Android
. In fact, some analysts believe that about half of the Android devices ... 阅读全帖
w**z
发帖数: 8232
9
来自主题: Programming版 - 请教goodbug等大神几个AWS的问题
Netflix 主要用Cassandra。 它的cross region replication是最好用的。只要setup
好,data 都能自动 replicate across data center。 它有现成的snitch
http://www.datastax.com/documentation/cassandra/2.0/cassandra/a
ELB 是用来distribute web traffic 到后台的instance。 autoscaling 是解决后代
instance 不够用的情况下,自动scale up
。等load 下来再scale down
不同region的 fail over, 主要是C* 做好来。其他都简单些。
c*******0
发帖数: 5247
10

replication,
using
binary
an
我看不出这个东西比Appengine好在哪里。Appengine也不用考虑replication,load
balancing,cache,有Android/iOS API,比parse便宜不知道多少倍。
当然你要写一点点后端程序,但有endpoint,这个一点点几乎可以忽略不计。
w**z
发帖数: 8232
11
来自主题: Programming版 - 问个master-slave的简单问题
strong consistency is very hard to achieve in distributed system. read below
for MySQL replication problem.
http://gtowey.blogspot.com/2012/10/how-can-mysql-replication-br
w***g
发帖数: 5958
12
来自主题: Programming版 - 有人上Spark用python API的么
所有模型都是周期性完整重算,还没到要用streaming的地步。
在线算法用C++写的,同时也负责处理上次模型计算以来新到的数据对模型的影响。
我觉得用C++写计算速度不是最主要的。肯定比scala快,但应该快不了好几倍。
C++的好处是省内存。100G内存可以处理全量用户。然后多台机器纯粹是
replicate增加吞吐量。用scala写在线模块的话,机器数量再加5倍估计都不够,
而且数据一旦出了内存,对吞吐量的影响是灾难性的。
我做的是音频类app的在线推荐,用户量从百度上查出来也有好几千万。同时在线的显
然没那么多。从性能和技术上来说,我可以用一台服务器搞定所有的推荐请求。实际肯定
不会那么做。但是replicate和sharding在实现难易程度上是有很大的差距的。
用C++使得整个在线服务的架构得到非常大的简化,架构简化得到的稳定性提高和维护
开销减少是非常大的。而且就是代码行数上来说,其实也并不吃亏。
离线用spark,看重的是可以很容易地尝试新的算法和模型。当时确实用的很爽。
g*****g
发帖数: 34805
13
Check Cassandra, Memcached, Redis design and you'll understand. The userId
can be hashed and distributed in a cluster. And any single piece of data can
be further replicated, consistency can be reached as long as read+write >
replication factor.
n*w
发帖数: 3393
14
any database replication, prefer rmdb transactional replication
g*****g
发帖数: 34805
15
来自主题: Programming版 - Cassandra Rewritten In C++, Ten Times Faster
http://www.scylladb.com/technology/cassandra-vs-scylla-benchmar
看了这个benchmark,目测10倍是没有,8倍?但关键是misleading,没有人会单机跑C*
的,你最少也上3个节点。我怀疑replication会由Network IO latency主导,从而大幅
缩小差距,有可能会然并卵。举个例子就是network latency 1ms, 你kernel bypass
节省了0.1ms,绝对是然并卵。
吹嘘的单机sharding优化没有意义,一个机器当了就全当了,用C*的目的之一就是追求
high availability. 一份数据多机备份是必须的,这反过来直接说明cluster里,
replication factor > 1立马要大打折扣。
在cluster benchmark出来之前,我对这东西保持怀疑态度。
n****s
发帖数: 119
16
真心请教.假如工作中遇到了类似12306的问题,要处理每秒一百万次的写操作,request
会发到四五个data center.
如果我需要strong consistency, 并且要cross datacenter replication,要怎么设计?
如果不需要strong consistency,也不要cross DC replication,要怎么设计?
单机版的就免了.难道每个数据中心放一台机器么?
F**********t
发帖数: 33
17
办公平台是Lotus Notes,要离开公司了,平时工作忙,多少年的邮件,绝大多数
attachments都没有下载。 平时也就是把邮件replicate到C盘,这样出差或者offline
的时候可以看。现在要走了,请问如何把这些资料copy到自己的盘里方便以后看,尤其
是那些attachments。我已经都replicate到C盘了,关键是以后怎么看,就像现在用公
司电脑一样,可以很方便的按照日期,发件人等检索。难道一定要用Lotus Notes吗?
好像很贵吧?这是一个大型办公平台,不像outlook. 个人怎么用?
请别误会,不是要盗取什么,只是以前的一些工作积累,不涉及机密。
多谢。
r**r
发帖数: 171
18
来自主题: Biology版 - MCBJC paper for discussion-042803

A Role for Ran-GTP and Crm1 in Blocking Re-Replication.
Yamaguchi R, Newport J.
Department of Biology, University of California, San Diego, 9500 Gilman Drive,
92093, La Jolla, CA, USA
All eukaryotic cells have regulatory mechanisms that limit genomic replication
to a single round each cell cycle. These systems function by blocking
formation of prereplication complexes. The regulatory mechanisms in the yeast
S. cerevisiae have been identified, but these do not appear to be conserved in
metazoan
M****e
发帖数: 70
19
来自主题: Biology版 - MCBJC paper for discussion-042803
this is a good paper that tries to dissect the regulatory
mechanism of strict control on DNA replication during cell
cycle. to ensure only one round of DNA replication at each
S phase, yeast S. cerevisiae exploits cdk2 kinase to regulate
functionally reduntant pathways: degradation of Cdc6, export
of MCM from the nucleus, and phosphorylation of ORC. however,
the mechanism in metazoan is unclear while the cellular
targets of cdk2 have not been completely understood yet.
the author (only one besid
d******u
发帖数: 178
20
每组的replicates数目。差异比较小的话要比较多的replicate才探测得到。
y******8
发帖数: 1764
21
来自主题: Biology版 - 问个一投稿的问题
One sample from google. Good luck!
Dear PLoS Biology editor:
I would like to submit a presubmission inquiry regarding suitability of our
manuscript, "Subversion of cellular autophagy pathway by RNA viruses" by
William T. Jackson, Thomas H. Giddings Jr., Sara Mulinyawe, Ron Kopito and
Karla Kirkegaard for publication in PLoS Biology. Cellular autophagy is a
field that is just entering a period of rapid discovery, because the
mammalian homologs for the many yeast genes that affect autophagy have v... 阅读全帖
o********r
发帖数: 775
22
这么说吧,这篇文章从初稿到接受花了11个月,对于BMC bioinformatics而言绝对不是
正常的速度,估计很多时间是花在说服reviewer那个1-to-1 comparison上面。为啥说
他是垃圾?因为这个前提就是错的,生物试验中的replicate是为了去掉一些偶然因素
对试验结果的影响,你们现在这么搞是反其道行之。举个例子,如果某个gene set里的
基因都受某个蛋白直接调控(比如TF),他们的mRNA level和这个蛋白的活性或者浓度
成正比,而且fold change特别大,但是这个蛋白和phenotype没关系,他的活性在
replicate中随机波动。如果是unpaired group analysis,这个set不会显著,但是用
你们的1-to-1 comparison,大多数情况下这个set都是highly significant,估计最后
这个set within top 10,而且不管用啥不同的试验条件都是highly consistency(本
来这个蛋白就是indepent的)。
你们用consistency来说明自己的方法好,但是没有证明你们找... 阅读全帖
B*******l
发帖数: 144
23
来自主题: Biology版 - 由鱼想到的(欢迎讨论)
I believe the polypoloidy cells you mentioned are from slavery gland.
The peri-centromeric heterochromatin of these cells are under replicated
based on cytological data, (I am not sure these regions are under replicated
or more compact). However these cells will not divide.

same
mutations which
formation
not
some
n***g
发帖数: 5027
24
这个总结不错:
申请精华帖:我了解的Epigenetics牛人一览
我是从C.D.Allis,Danny Reinberg,Tony Kouzarides,Yi Zhang, Yang Shi这几个
人发的文章看起的,这几个人基本上就是目前做表观遗传最牛的一批人了。然后你看看
他们文章citation里面出现频率比较高的人名,那往往就是大牛了,然后一点一点扩展
。个人感觉看CNS及其子刊还是王道。另外推荐一个很好的网站,http://www.epidna.com/ 可以查这个领域最top的lab,也可以查该领域最新的论文。
# ^2 P/ }, _0 h" U8 m) P6 b
; l* o$ y- |- V5 g0 I0 T然后在申请选校的过程中,我也逐渐了解到一些
epigenetics领域的big names和rising stars。所以我感觉看各个牛校的faculty也是
了解一个领域的好学校好lab的方法。我的做法是每看一个学校,就把相关领域做
epigenetics的faculty留下来,然后以后看CNS时或者Journal club上看到熟悉的名字
,就重点关注下。$ ... 阅读全帖
n********k
发帖数: 2818
25
As you folks are discussing Cori, I happened to read his autobiography and
find it both interesting and enlightening:)))...BTW, it sounds Richard Axel
is a lot nicer than what I/we have heard...or in another word, his cruelty
is rooted in his obsession with science and indifference to humanity:))); As
well, it is pretty...he noted Cori but left his ex out...pretty human or
not human?
Autobiography
Richard AxelNew York City is my world. I was born in Brooklyn, the first
child of immigrant parents... 阅读全帖
w******e
发帖数: 1187
26
A system for the continuous directed evolution of
biomolecules
Kevin M. Esvelt1, Jacob C. Carlson2 & David R. Liu2,3
Laboratory evolution has generated many biomolecules with
desired properties, but a single round of mutation, gene expression,
screening or selection, and replication typically requires days or
longer with frequent human intervention1. Because evolutionary
success is dependent on the total number of rounds performed2, a
means of performing laboratory evolution continuously and
rap... 阅读全帖
p*****m
发帖数: 7030
27
所谓重复 没人规定一定要是不同时间反复做吧。。。你们现在lab如果一次能把多个
Biological replicates和technical replicates都做了 不是挺好的么

天。
p*****m
发帖数: 7030
28
他们收集Biological replicates,用到的东西已经不一样了吧。当然你可以认为一切环
境因素都有可能干扰结果 比如说冰箱温度 天气 湿度 你说话声音 咳嗽了没有 试剂批
次 blabla 这样重复起来就没完没了了。。你就是做100次也不可能穷举所有的变量。
打个比方 做deepseq或者microarray 有人是先后run好几批然后分析的么?大家都是收
集几个bioligical replicates一起run的啊
a***y
发帖数: 19743
29
来自主题: Biology版 - 关于Error bar的问题
这篇文章讨论了error bar的问题
Error bars in experimental biology
Geoff Cumming,1 Fiona Fidler,1 and David L. Vaux2
独立试验做一次,算n=1,不管你里面有多少biological replicates, technical
replicates。
n越大越好。所以应该是另肯bio和tech的rep少一点,也要把试验多重复几次,n才会上
去。
但是如果一个试验做一次要两个月,重复两次,一共花半年,n才等于三。对于一些系
统误较大的试验,实验与实验之间差异较大,但是单个实验内的差别,趋势都是很明显
,但是最后n=3来做出mean和error bar然后做statistics,会不significant,那怎么
办呢?
因为我知道一些领域实验不重复的,有些是因为重复的话成本太高。
还有一些领域会展示三次重复实验中的一次的结果,给出SE或SD然后还算p value,说
三次实验的结果,趋势都类似。
这个按照这篇文章貌似是不可以的。
那么到底应该怎么做?一个实验耗时很长,每个实验之间系统误差较大的,咋办... 阅读全帖
s******s
发帖数: 13035
30
来自主题: Biology版 - 关于Error bar的问题
阿提又出来调皮了。
都不用看那篇文章,就知道你总结的不对。
独立实验的replication,是保证多次实验的一致性; 同一次实验的replicate
是保证这次实验里面的一致性,N当然也是有用的。生物实验一般是有control,
也就是对照组的,只要每次实验都有自己的对照,那么毫无必要追求多次实验
的数据一致,只要结论一致就行了。
至于tech rep, 那是凌驾于多次实验之上的。
L*******a
发帖数: 293
31
来自主题: Biology版 - NGS 二代测序分析,大家来评评
Generally, 70% mappable reads is an experienced criteria.
FYI.
ENCODE project requires 10M mappable reads for human ChIP-seq experiments,
and modENCODE project requires 4M for worm and fly.
From the perspective of genomic, some classic genes bound by peaks are not
enough to prove your ChIP-seq data is valid, in terms of saturation.
You'd better to have some replicates anyway, each of replicate should meet
saturation and mappability criteria, and the reproducibility rate also
should be in an acce... 阅读全帖
j****x
发帖数: 1704
32
成本,成本,还是成本
实验部分中最花钱的就是建库,测序成本其实相对很低尤其是有了barcode,按照你的
设想做出来的technique replicates,人家不如直接做biological replicates了,不
是更有意义?
a***y
发帖数: 19743
33
别人的实验数据帮分析。
有三个replicate的control,用来作为基数算实验组fold change的
有些时间point,replicate之间Ct差异较大。如果efficiency是1,1 Ct是大约2倍,那
么有些时间段内
9个(3 x 3)Ct值的range差异可以有常常大于2-3,有个时间段range有4-5。
这个control还能用么?这个实验数据还能发表么?
相比之下各个实验组内的数据Ct差异都很小。
我感觉这个control差异这么大,作为基数很problematic,而且差异这么大的Ct值做算
术平均数距离真实值差异也太大了,必须换一个算法。(以前版上有人求excel表格,
还有没有啊?)
谢谢了!发包子感谢回答的。
a***y
发帖数: 19743
34
谢谢。给你发包子了。
补充说明一下
这个是三个biological replicate
然后跑qPCR的时候有每个biological replicate有三个tech rep
qPCR tech rep之间差异很小,所以不是loading的问题
biological rep之间Ct差异很大。作为control,我觉得有这么大的差异,使得数据分
析和结果分析几乎不可能……除非接受平均值是可信的。
还有谁能给点意见?
谢谢!
n****0
发帖数: 696
35
来自主题: Biology版 - RNA 260/230 ratio一问
做过了,三个在同一板上的technical replicates还可以,如果Ct能在30以下的话。不
过我的确发现有些边缘位置的孔总是和另两个不一样。但是biological replicates就
不行了,同样样品做两次也不一样。这次差5倍,下次差10倍 (relative fold change
),甚至有时趋势还不一样。可能是我提取RNA的问题。因为我要分离tissue,取样品
时间比较长,虽然我都是尽量快速的放到液氮里。几分钟的时间,转录水平会有很大变
化么?
F*Q
发帖数: 3259
36
来自主题: Biology版 - 中医还是很牛的!
中医的以毒攻毒在治疗癌症方面的应用!
From New York Times
Viruses Recruited as Killers of Tumors
By RACHEL NUWER
Published: March 19, 2012
THE HYPOTHESIS
Viruses can be designed to attack tumors.
THE INVESTIGATORS
Dr. Robert Martuza, Harvard Medical School; Dr. Bernard Roizman, University
of Chicago; Dr. Ian Mohr, New York University.
In 1951, a 4-year-old boy with leukemia contracted chickenpox. His liver and
spleen, swollen by the cancer, soon returned to normal, and his elevated
blood cell count fell to that of... 阅读全帖
r*****t
发帖数: 4793
37
来自主题: Biology版 - paper help!
Nature. 2012 Mar 14;483(7390):434-8. doi: 10.1038/nature10895.
Intrinsic coupling of lagging-strand synthesis to chromatin assembly.
Smith DJ, Whitehouse I.
Source
Molecular Biology Program, Memorial Sloan-Kettering Cancer Center, 1275 York
Avenue, New York, New York 10065, USA.
Abstract
Fifty per cent of the genome is discontinuously replicated on the lagging
strand as Okazaki fragments. Eukaryotic Okazaki fragments remain poorly
characterized and, because nucleosomes are rapidly deposited on n... 阅读全帖
d****d
发帖数: 214
38
Some insightful comments on the article:
JJ Bangor, ME
In years past, Nature, Science and the Cell Press Journals would devote
maximal energy to make sure scientific accuracy was maintained for all
articles that were published on their pages. Sadly, this is no longer the
case. In fact, a journal's impact factor is only helped by bad papers that
attract a lot of attention and are later found out to be fraudulent or
simply wrong. This is the case simply because these papers continue to be
cited... 阅读全帖
h*******o
发帖数: 4884
39
The outcome will largely be the same.
As the new cybrids replicate with the infused mitochondria, the replicated
mitochondria will contain your target mtDNA, proteins encoded by your target
DNA and proteins encoded by your original nuclear DNA but not the exogenous nDNA.
l**********1
发帖数: 5204
40
antagonisctic co-evoultionary Replicators 吧
由 Kin selection Red Queen Dynamics 定输赢
one PhD thesis (2009) from Barcelona
by Cayuela JS
title:
Dynamics, Evolution and Inoformation in Nonlinear Dynamical Systems of
Replicators
PDF full text link:
//www.tesisenred.net/bitstream/handle/10803/7182/tjsc.pdf?sequence=1
D*a
发帖数: 6830
41
来自主题: Biology版 - Nature 对 Science 的逆袭
Nature忘了它有一整本杂志都是这玩意儿了?
两年前Science发了一篇GWAS方法研究人类长寿相关基因分析的文章,一年后被说有技
术问题,假阳性,撤稿了,最近刚从Plos One发出来了。两年前就瞅了一眼结果,现在
看貌似大结论没变啊?还没细看,细看也看不懂。。。
http://www.nature.com/nature/journal/v487/n7408/full/487427a.ht
Scientists and journals must work together to ensure that eye-catching
artefacts are not trumpeted as genomic insights, says Daniel MacArthur.
When a study of the genomes of centenarians reported genetic variants
strongly associated with exceptional longevity1, it received widespread
media and publi... 阅读全帖
s**********t
发帖数: 680
42
来自主题: Biology版 - Paper request! Thanks!
1.
DNA Replication and Nuclear Architecture
David M. Gilbert
Susan M. Gasser
in "DNA Replication and Human Disease" (2006, Volume 47)
Cold Spring Harbor Monograph
DOI: 10.1101/087969766.47.175
Link: https://cshmonographs.org/index.php/monographs/article/view/4740
2.
Structural-functional organization of chromosomal DNA domains
Razin SV, Hancock R, Iarovaia O, Westergaard O, Gromova I, Georgiev GP.
Cold Spring Harb Symp Quant Biol. 1993;58:25-35.
Link: http://symposium.cshlp.org/content/58/25.ext... 阅读全帖
s**********t
发帖数: 680
43
来自主题: Biology版 - Paper help!
Title: Eukaryotic Replication Origins and Initiation of DNA Replication
Melvin L DePamphilis,
Mirit I Aladjem
Encyclopedia of Life Sciences
Published Online: 15 SEP 2010
Link: http://onlinelibrary.wiley.com/doi/10.1002/9780470015902.a0001055.pub2/abstract
DOI: 10.1002/9780470015902.a0001055.pub2
Please send it to: b************[email protected]
Thank you very much for your help!
D*a
发帖数: 6830
44
来自主题: Biology版 - 癌症是生命进化的伟大成就
酵母有两种寿命,replicative lifespan 和 chronological lifespan,不能分裂了表
示replicative lifespan 终止,但是酵母在不能分裂了之后还能活一段时间,有点类
似人的衰老和生殖。
对单个cell来说端粒很重要,但是很多生物,老年体的细胞提取出来还能分裂十几次还
是几十次,说明细胞分裂能力并不是最终诱因,包括干细胞的保持,不光跟本身分裂能
力有关,跟其所处的niche也有很大关系。
包括c elegans里面如果过表达telomere-binding protein HRP-1可以延长端粒和寿命
,但是因为线虫体细胞不再分裂,所以端粒酶很可能确实是对衰老有影响,但是应该是
通过其他途径(比如oxidative stress神马的,paper没有细看)而不是通过维持细胞
的正常分裂能力来影响衰老。如果承认衰老是进化上普遍存在的现象,c elegans这个
例子就很有说服力。
老鼠里面有两篇关于telomerase改善老鼠衰老的症状等等的文章一篇在cancer
resistant的模型里面(否则telomerase让老鼠得各种癌症早... 阅读全帖
c***3
发帖数: 251
45
来自主题: Biology版 - 问个统计的问题
我觉得他问Standard deviation的意思,主要就是看你这些replicates之间是不是有很
大差异。
你是可以求SD,问题你一共有几个replicates啊?如果你有几十个,画一个boxplot好
s******s
发帖数: 13035
46
来自主题: Biology版 - 问个统计的问题
chi-square是一个很垃圾的test, 之所以有人用完全是历史遗留问题。
至于是不是用t-test,看你这个实验每次replicate理论上会不会有大区别了。
如果要显示这儿区别,当然要t-test;基本上没有,其实根本不用replicate,
直接把每个点看做bernoulli trial, 然后总得是binomial, 直接用CLT估计误差就行了
j*p
发帖数: 411
47
来自主题: Biology版 - RNA-seq 表达量问题
1. RPKM = 1 约等于 1 copy/cell. 同样是rpkm=1,如果这个是从100M reads出来的,
可信度比从10M来的高。同时,RPKM=1可以通过单细胞FISH验证。
2. qPCR 灵敏度比100M RNAseq高,能够validate RPKM=0.1左右,就是cycle要多些。
3. 即使没有replicate,也可以做统计,cufflinks, DESeq 都有这样的选项(简单的
fisher-exact test),但得出的pvalue显然没有那些有replicate的来得靠谱。
4. 做表达量的时候,通常会用 log2(RPKM+1),然后做fold change的时候,会用log2
的差,+1既是为了去0,避免fold change = 无穷大,也是为了减少对那些表达量很小
的RNA的fold change的over estimation。
首页 上页 1 2 3 4 5 6 7 8 9 10 下页 末页 (共10页)