文档库 最新最全的文档下载
当前位置:文档库 › Massively parallel BLAST for the Blue GeneL

Massively parallel BLAST for the Blue GeneL

Massively parallel BLAST for the Blue GeneL
Massively parallel BLAST for the Blue GeneL

1

Massively Parallel BLAST for the Blue Gene/L

Huzefa Rangwala,Eric Lantz,

Roy Musselman,Kurt Pinnow,Brian Smith and Brian Wallenfelt

Computer Science Department,University of Minnesota-Twin Cities,MN55455

Email:rangwala@https://www.wendangku.net/doc/8715200024.html,

Computer Sciences Department,University of Wisconsin-Madison,WI53706

Email:lantz@https://www.wendangku.net/doc/8715200024.html,

IBM Systems and Technology Group,Rochester,MN55901

Email:mussel,kwp,smithbr,bwallen@https://www.wendangku.net/doc/8715200024.html,

Abstract—The focus of this article is to explain our research involved with running a parallel implementation of the widely used BLAST algorithm on thousands of processors,available on supercomputers like the IBM Blue Gene/L.Our work involved optimally splitting up the set of queries as well as the database. We also found solutions to reduce the I/O thereby delivering a fast,high throughput BLAST.Our results show that we are capable of performing at least2million BLAST searches per day against a database of2.5million protein sequences.We also show very good performance in terms of speedup and ef?ciency.

I.I NTRODUCTION

There has been an ever increasing growth in the size of nu-cleotide and protein databases due to advances in sequencing technology.There is a high demand in the area of computa-tional biology to extract useful information from these massive databases.This has led to the usage and development of high performance computing power to help researchers sift through the voluminous biological data.Supercomputers,clusters,and custom designed bioinformatics hardware solutions attempt to tackle the various problems stemming from analysis of biological data.

BLAST(Basic Local Alignment Search Tool)[1],[2]is a widely used bioinformatics application for rapidly querying nucleotide(DNA)and protein sequence databases.Given a query sequence,the goal is to?nd the most similar sequences from a large database.This has applications in the identi?-cation of functions and structures of unknown sequences or understanding the evolutionary origin of DNA or protein query sequences.Due to its high importance,BLAST also?nds a place in almost all the bioinformatics solutions.

The simplest approach to?nd the most similar database sequences to a query is to calculate the pairwise score of the alignment between the query and every sequence in the database,using an alignment algorithm like Smith-Waterman. However,the computational cost of this method is high,and most of the alignments have very poor scores.BLAST uses some heuristic methods to reduce the running time with little sacri?ce in accuracy.BLAST is available on the web or for download and the most popular implementation is from Note:Huzefa Rangwala and Eric Lantz both shared equal portion of the ideas and work presented in this paper.They should both be considered as the?rst authors of this work.NCBI(the National Center for Biotechnology Information, https://www.wendangku.net/doc/8715200024.html,/BLAST).

Recently,there have been several approaches to using BLAST on supercomputers and clusters.Some of these ap-proaches[3],[4],[5]focus on the“embarrassingly parallel”solution of distributing the query set across several cluster nodes,each of which executes a serial job.Throughput is increased,but the time for a particular query to complete is un-changed.Another approach is to parition the database among cluster nodes and have each node search an assigned part of the database for the same query[6],[7],[8].This approach of database splitting was developed in the mpiBLAST[6] implementation and optimized in pioBLAST[9].Only one query is searched at a time,but the portion of the database each processor has to look at is reduced.

Our research focused on using a larger number of processors for running BLAST than ever before.This was done with Blue Gene/L[10],the world’s fastest supercomputer1.Each refrigerator-sized rack of Blue Gene/L consists of1024nodes. Each node has two700MHz PowerPC440processors and512 MB of https://www.wendangku.net/doc/8715200024.html,pared to other large supercomputers,the Blue Gene has low power consumption per processor and a very fast interconnect network.

Our parallel BLAST solution not only focused on the database fragmentation as introduced in mpiBLAST but also integrated query load division.This scheme allowed us to use a large number of processors.We introuced new input and output handling schemes extending the ef?cient data access methods seen in the pioBLAST implementation.

In this report we begin Section II with a discussion of related work in terms of mpiBLAST and pioBLAST.We follow this with the discussion of our BLAST solution as well as introduced I/O techniques in Section III.Section IV shows our results on the Blue Gene/L for various datasets.We close the paper with Section V which states the conclusions and future directions for this area.

II.P REVIOUS W ORK

Earlier work in achieving a faster BLAST solution has led to customized BLAST hardware[11],[12],shared memory parallel BLAST and distributed memory parallel BLAST.As 1https://www.wendangku.net/doc/8715200024.html,

2

discussed in[9]there have been two adopted techniques:one dealing with splitting of the database and the other dealing with the query load distribution.In this section we discuss the fairly recent work using the database fragmentation strategy.

A.mpiBLAST

mpiBLAST[6]was developed at Los Alamos National Laboratories.This work introduced the database fragmentation strategy in the context of BLAST.It is an open source project that uses the standard message passing protocol(MPI)[13]for its parallel BLAST implementation.It works on a wide range of clusters and supercomputers and has gained popularity amongst members of the bioinformatics community needing a high performance BLAST.

It works by initially dividing up the database into multiple fragments so that each processor has a separate smaller frag-ment to work on(this program is called mpiformatdb).The searching of a fragment is independent of any other fragment lending a very parallelizable solution.

Most of the nodes are workers:they open up a fragment?le, call the NCBI serial code to conduct the search,and return the results.There is one scheduler node that reads in the query (or queries),broadcasts to the other nodes,and decides which fragments each worker should search.The last node is a writer node that receives the results from each worker,sorts them by score,and writes the output?le.

mpiBLAST tries to substantially reduce the disk I/O read times by reducing the size of database to be read in by each cluster node and making it small enough to?t in memory. However,mpiBLAST suffers from non-search overheads with increasing number of processors and varying database sizes as discussed in[9].

B.pioBLAST

In a data driven application like BLAST,it is apparent that poor handling of I/O can lead to performance bottlenecks. pioBLAST[9]optimized mpiBLAST by introducing?exible database partitioning options,enabling concurrent access to shared?les by use of parallel I/O,and caching of intermediate ?les and results.

pioBLAST stands for parallel I/O BLAST and uses MPI-IO[14],[15]for ef?cient data access.MPI-IO allows mul-tiple processors to read or write?les simultaneously.This is especially helpful in combining the results from all of the processors at the end of the run.The scheduler and writer nodes are combined into a single process called the master.The database fragments and other?les are mapped into memory buffers,so they are not read more than once, even through multiple queries.

pioBLAST is also able to segment the database dynami-cally,eliminating the need to preformat the database.However, this strategy would not work well when the database size exceeds the memory on the master node.We did not test this functionality as it was ef?cient to have a distributed database, fragmented beforehand.This would allow us to ef?ciently read the entire database through multiple sources as described later in Section III.

When we ran pioBLAST for increasing number of frag-ments,we noticed that the running time no longer improved signi?cantly.There was a limit to how many fragments the database could be broken into.This was due to the fact that as the number of fragments grew the amount of time needed to combine the results increased,resulting in a higher cost for parallelization.

III.M ETHODS

To utilize as many nodes of the Blue Gene/L as possible, we implemented an additional parallelization of distributing the query set along with the database segmentation.We were also aware of the strengths of the Blue Gene/L,a large number of low power processors integrated with high speed interconnection networks.The work described in[16]gave a good overview of designing scalable applications for the Blue Gene/L.

We de?ned a group as a set of processors working on a fraction of the total input queries,with each processor having a fragment of the database and the group collectively having the entire database.Each group consists of a master node and a number of workers.For our tests,we use what the pioBLAST paper[9],calls natural partitioning,where the number of processors in the group is one more than the number of database fragments.Each group is assigned a subset of the overall query sequences.For a complete job,the overall number of processors used is given by the equation:

(1) The current implementation only supports static query set partitioning.Therefore we needed to attempt to distribute the query set so that each group would have approximately the same work load.We found that balancing the total lengths of the queries assigned to each group(as opposed to the number of queries)worked well to avoid processor idling.

Rather than having all the groups reading the database from the input source,we employed an ef?cient scheme where only the?rst group was assigned to be the reader.The processors in the reader group would be responsible for reading the needed fragment.These processors would then broadcast the fragment to corresponding processors needing the same fragment in the other groups.This scheme,which we call the broadcast strategy,is illustrated with Figure1,where the red line shows the ef?cient data?ow of this strategy.The blue line represents the alternative all-reader strategy,where every processor must retrieve a fragment from disk.

Instead of using a single disk as the input source we replicated our database on multiple?le servers(four IBM X345systems).The database read load was shared across the four?le systems.This method also distributed the load of writing intermediate?les.This technique is similar in concept to data mirroring and has been widely used before to reduce disk access bottlenecks in other applications[17].

We also employed a radical model of driving input data into the application by implementing a server agent.This agent

3

n F r a g m e n t s

Data Flow for Non Broadcast Scheme (All processors reading from the I/O system) Data Flow for Broadcast Scheme

Fig.1.

Parallel Blast Model.The database is read from one of the three sources shown on the left via one of the data ?ow schemes.

would store the database fragments into its memory and then transfer the data to the worker nodes when requested using the TCP connection to avoid the ?le system overhead.Currently,our implementation used a single server and single port to handle the incoming requests for database fragments from the worker nodes.The server process would spawn child processes depending upon the fragment requested and set up a new port for the transfer.This strategy was an initial model and showed promising results.

We saw an increase in the performance when we replicated the servers such that the requests made by the worker nodes was distributed across the several servers.In the future we would like to try a dedicated server-fragment strategy where a server would be responsible for only one fragment.The worker nodes would have information regarding the server they would need to connect for getting their fragment.From our single server model results,we anticipate a substantial reduction in the input read times with the dedicated server-fragment strategy.

IV.P ERFORMANCE E VALUATION

In this section we evaluate our parallel BLAST solution on the Blue Gene/L.We ran several experiments with the different techniques proposed in the earlier section.This helped us in understanding the optimal parameters for achieving a high throughput and highly scalable parallel BLAST implementation.Most of the tests were performed on the development Blue Gene/L system at IBM Rochester.This development system consisted of 4096nodes,where each node consisted of two 700Mhz PowerPC 440d processors.The system was enclosed in four racks or towers.

A.Datasets

For benchmarking purposes,we used freely available datasets and databases.There were two databases and three input ?les used in the data reported in this paper.One problem that we faced when benchmarking or performing comparative studies was that the database were continually updated with new sequences.Databases:

1)nr:The nonredundant protein database.Contains 2.5million protein sequences as of June 2005. 1.2GB unformatted,1.6GB formatted.Source:NCBI 2)nt:Nucleotide database.Contains 2.5million nucleotide sequences as of August 2005.16GB unformatted,4.5GB formatted.Source:NCBI Query ?les:

1)arabidopsis:1168EST protein sequences from Arabidopsis thalinia .Average length:465bases (min 128,max 706).Source:Bioinformatics Benchmark System v.3from Scalable Informatics 22)arabidopsis full:28014EST protein sequences from Arabidopsis thalinia .Average length 419bases (min 28,

2The

?le has “tomato”in its name,but the sequences all have arabidopsis

in their descriptions

4

max 715).Source:Bioinformatics Benchmark System v.3

3) e.chrysanthemi:441Erwinia chrysanthemi bacterial genome nucleotide sequences.Average length 658bases (min 96,max 4367).Source:Aaron Darling (used in [6])All of the graphs and tables reported in this paper benchmark the performance of blastx (querying a nucleotide sequence against a protein database),using the nucleotide arabidopsis query ?le against the protein nr database.The performance results remain similar when using blastn (querying a nu-cleotide sequence against a nucleotide database)or blastp (querying a protein sequence against a protein database).Additional results are from blastx query of arabidopsis full versus nr and blastn query of e.chrysanthemi versus nt .B.Results

A signi?cant portion of the non-search fraction of the total BLAST run time is dependent on the database read time.Reading a large database can cause a huge stress on the network bandwidth as well as the I/O of any system.As mentioned in Section III we employed various strategies to reduce the database read time compared to the actual search time.We ran several experiments to compare these database reading techniques using the arabidopsis dataset against the nr database.

Figure 2shows the database read times using 63fragments with an increasing number of processors for various data driving strategies.Both the single disk (scratch)and the multiple X345?le systems show improvement with the broad-cast strategy over the all-reader strategy.The data mirroring technique as seen with the use of four X345systems gives the best I/O performance which ?nally gets re?ected in the total run time.This is illustrated in Figure 3,which shows the total run times with increasing number of processors using 63fragments.It was interesting to observe that with the broadcast strategy we were able to keep the input read times constant

20 40 60 80 100 120 140 160 180 0 200 400 600 800

1000 1200 1400 1600 1800 2000 2200T i m i n g (s e c o n d s )

Number of Processors

I/O Time(Without Bcast) Scratch Disk I/O Time (With Bcast) Scratch Disk

I/O Time(Without Bcast) X345I/O Time (With Bcast) X345Fig.2.Database Read Time Performance for x345vs Scratch Disk.The database was split into 63fragments.as we increased the number of groups and hence the number of processors.

TABLE I

Percentage improvement in the database read times using single group read

and broadcast strategy.Number of fragments used =31Scratch Disk

X345Systems

Groups No Bcast Bcast Improv.

No Bcast Bcast Improv.19.279.270.00%8.008.000.00%218.619.4949.01% 5.808.00-37.9%421.969.8555.15%9.547.9316.88%849.6311.0477.76%23.699.9258.13%16115.1411.6189.92%28.0814.7947.33%32220.6215.6692.90%39.889.8875.23%64

311.3126.44

92.00%

77.2615.2980.21%

The percentage improvement in database read times is

shown in Table I.The percentage improvement of the broad-cast strategy over the all-reader strategy rises considerably with an increasing number of processors.As we increase the number of processors,the all-reader strategy causes concurrent access to the input source and leads to congestion.However,in the broadcast strategy there are a constant number of pro-cessors accessing the input source irrespective of the number of groups.Hence,we see a constant read time for the same number of fragments which explains the high performance gain.

We performed a similar study on the database read times for our simple implementation of single server single port model.It was observed that the database read times were comparable to a single disk results.Preliminary testing of replicating servers,such that multiple servers would share the load as input sources showed relatively better performance compared to a single server.The next step would be to test the dedicated server-fragment strategy described earlier in section III.

The other main component of our experiments was to establish the optimal number of groups and fragments that could be run on the Blue Gene/L utilizing as many processors as possible.Figure 4shows the total run time across different partition sizes,using increasing number of groups.Employing the broadcast strategy,along with a single input source,we

200 400 600 800 1000 1200 1400 1600 1800 2000 0 200 400 600 800

1000 1200 1400 1600 1800 2000 2200

T i m i n g (s e c o n d s )

Number of Processors

Without Bcast - Scratch Disk

With Bcast - Scratch Disk

Without Bcast - X345With Bcast - X345

Fig.3.Blast Run Time Performance for x345vs Scratch Disk.The database was split into 63fragments.

5 Fig.4.Total runtimes for using different partition sizes with varying number of processors on a single scratch disk.For example,following equation1,the

tallest bars represent2groups of15fragments and1group of31fragments,respectively.

see that there is almost perfect scalability as we increase the

number of groups for almost all fragment sizes.It is also

interesting to note that the run time remains the same for same

number of processors with different fragment size.

However,for511fragments increasing the number of pro-

cessors leads to an increase in the run times.Looking at the

timing pro?les,we realized this was due to the large number

of intermediate?les and results that needed to be recombined.

The size of the fragment to be searched per processor becomes

very small,making the results combination step greater than the actual search time.

Figure5shows the speedup performance achieved by our scheme with increasing processors across different fragment sizes.Unlike mpiBLAST results[6],we assume our serial run times to be void of any paging overhead and hence compute it with reference to a parallel job using16processors on the Blue Gene/L.

For the same arabidopsis dataset,the running time on16 processors was over7000seconds,which reduced almost lin-early to only73seconds(our best run time for the benchmark) using a full rack(1024nodes/2048processors,32groups of64processors).A larger test with28,014EST sequences (arabidopsis full)?nished in under20minutes,equating to an amazing rate of2million sequences per day.These runs were made using the four X345systems and the broadcast strategy. We also compared our results against the e.chrysanthemi benchmark used by the mpiBLAST group3on the Green Destiny cluster[18].We were able to complete the task in71 seconds using512nodes of Blue Gene.The reported run time for mpiBLAST on the Green Destiny cluster was10minutes using128processors.

3https://www.wendangku.net/doc/8715200024.html,/About.Performance.html

100

1000

100 1000 S

p

e

e

d

u

p

T

s

/

T

p

[

l

o

g

s

c

a

l

e

]

Number of Processors [log scale]

Fig.5.Speedup curve on scratch disk with varying fragment sizes

V.C ONCLUSIONS

Our work has helped achieve a massively parallel BLAST search engine,delivering high throughput while remaining ef?cient with the usage of a large number of processors.The result of this work is what we believe to be the fastest BLAST machine yet developed.Our tests show it is able to handle many times more queries per day than NCBI4,though we were have not yet tested as diverse a set of queries and databases as available on NCBI’s servers.

Using thousands of processors raises technical challenges not conceivable on the level of a few dozen processors, 4As of April2005,NCBI handles400,000queries per day according to ACM Queue vol.3,no.3,April2005

6

and issues such as I/O and problem decomposition must be resolved in even more detail.We have introduced several practical ways of reducing the disk access times in data driven applications like BLAST.

Currently,our query partitioning technique is static and depends on the number of characters to be processed by each group.Dynamic load balancing and partitioning of the query set would be a possible future study.It would also be interesting to test different algorithms for distribution of the query set as well as database fragmentation.

We also discussed the possible implementation of a dedicated-fragment server model for better handling of the re-quests made by the several worker nodes.Future opportunities for research include adapting this structure for non-heuristic homology searches using the Smith-Waterman algorithm.An-other widely used version of BLAST called PSI-BLAST iter-atively re?nes the searches.Parallelizing this implementation would involve challenges of trying to predict the sequences that would be used for the next round of search.The ideas in this paper could also be extended to several other data driven applications on massively parallel environments.

A CKNOWLEDGMENTS

The work was supported by the IBM Systems Technology Group at Rochester,Minnesota.Huzefa Rangwala and Eric Lantz were working as interns and would like to express deep gratitude to all the members involved with the Blue Gene/L’s software and development team.Special thanks to our manager Patrick Keane,technical leaders,Jose Moreira, Jeff Parker,Tim Mullins,Carlos Sosa,Charles Archer and Joseph Ratterman(all involved with the Blue Gene/L Systems Team)for their encouragement and technical help in this effort. We would like to express our gratitude to Heshan Lin of NCSU for helping us with the pioBLAST code.

IBM,PowerPC,and Blue Gene are registered trademarks of IBM Coorporation.

R EFERENCES

[1]S.F.Altschul,W.Gish,https://www.wendangku.net/doc/8715200024.html,ler,E.W.Myers,and D.J.Lipman,

“Basic local alignment search tool,”Journal of Molecular Biology,vol.

215,no.3,pp.403–4010,1990.

[2]S. F.Altschul,L.T.Madden, A. A.Schffer,J.Zhang,Z.Zhang,

https://www.wendangku.net/doc/8715200024.html,ler,and D.J.Lipman,“Gapped blast and psi-blast:a new gen-eration of protein database search programs,”Nucleic Acids Research, vol.25,no.17,pp.3389–402,1997.

[3]R.Braun,K.Pedretti,T.Casavant,T.Scheetz,C.Birkett,and C.Roberts,

“Parallelization of local blast service on workstation clusters,”Future Generation Computer Systems,no.6,2001.

[4]N.Camp,H.Cofer,and R.Gomperts,“High-throughput

blast,”https://www.wendangku.net/doc/8715200024.html,/industries/sciences/chembio/resources/ papers/HTBlast/HT Whitepaper.html,1998.

[5] E.Chi,E.Shoop,J.Carlis,E.Retzel,and J.Riedl,“Ef?ciency of

shared-memory multiprocessors for a genetic sequence similarity search algorithm,”Technical Report,University of Minnesota,CS department, vol.TR97-05,1997.

[6] A.E.Darling,L.Carey,and W.Feng,“The design,implementation,and

evaluation of mpiblast,”In proceedings of CluterWorld and Expo,2003.

[7]R.Bjornson, A.Sherman,S.Weston,N.Willard,and J.Wing,

“Turboblast(r):A parallel implementation of blast built on turbohub,”

Proceedings of the International Parallel and Distributed Processing Symposium,2002.

[8] D.Mathog,“Parallel blast on split databases,”Bioinformatics,vol.19,

no.14,2003.

[9]H.Lin,X.Ma,P.Chandramohan,A.Geist,and N.Samatova,“Ef?cient

data access for parallel blast,”International Parallel and Distributed Processing Symposium,2005.

[10]“Blue gene,”IBM Journal of Research and Development,vol.49,no.

2–3,2005.

[11]R.Luthy and C.Hoover,“Hardware and software systems for acceler-

ating common bioinformatics sequence analysis algorithms,”Biosilico, vol.2,no.1,2004.

[12]K.Muriki,K.Underwood,and R.Sass,“Rc-blast:Towards an open

source hardware implementation,”Proceedings of4th IEEE Interna-tional Workshop on High Performance Computational Biology,2005.

[13]“Mpi:Message-passing interface standard,”Message Passing Interface

Forum,1995.

[14]J.May,“Parallel i/o for high performance computing,”Morgan Kauf-

mann Publishers,2001.

[15]R.Thakur,W.Gropp,and E.Lusk,“On implementing mpi-io portably

and with high performance,”Proceedings of the6th Workshop on I/O in Parallel and Distributed Systems,1999.

[16]G.Almasi and et.al,“Early experience with scienti?c applications on

the blue gene/l supercomputer,”Proceedings of the11th International Euro-Par Conference,Euro-Par2005,vol.3648,2005.

[17] C.Nyberg,T.Barclay,Z.Cvetanovic,J.Gray,and D.Lomet,“Alphasort:

A cache-sensitive parallel external sort,”VLDB,vol.4,pp.603–627,

1995.

[18]W.Feng,“Greendestiny+mpiblast=bioinfomagic,”Proceedings of the

10th International Conference on Parallel Computing,2003.

伺服驱动器外部接线及详细说明

一、伺服驱动器外部接线及说明。 SRV-CTRL PLC 外部电源 (DC 24V) SRV-CTRL 说明: 1、SRV-CTRL指PLC的伺服控制模块; 2、以→指向者为输入,以←指向者为输出。 3、各信号含义如下: INH: 禁止输入脉冲指令(开路时禁止);OZ-: Z相输出;OZ+: Z相输出; S-RDY: 伺服驱动器已准备就绪;ALM: 伺服驱动器故障报警; COIN: 定位已完成;SRV-ON: 伺服驱动器“开”信号; COM-: 电源负极;COM+: 电源正极; PULS1: 指令脉冲输入端; PULS2: 指令脉冲输入端; SIGN1: 指令脉冲符号输入端; SIGN2: 指令脉冲符号输入端; 二、参数说明: 1、参数设置方法。

操作面板上共有5个按键,意义如下: MODE:模式转换键,按此键可在4个模式间切换,这4个模式是: DP-××××: 选择监视项目(共有7个, 在按 MODE键显示DP-××××后先按SET,再按↑或↓选择) ⑴、DP-EPS: 位置偏差;⑵、DP-SPD: 转速; ⑶、DP-TRQ: 转矩;⑷、DP-CNT: 控制方式; ⑸、DP-IO: 输入输出信号状态;⑹、DP-ERR: 错误信息; ⑺、DP_NO: PR-××××: 设定参数。 EE-××××: 写入参数。 AT-××××: 自动增益调整。 SET:为设定及确认键。 ↑:数值增加或移动到下一个选项; ↓:数值减少或移动到上一个选项; ←:数位间移动; 具体设置步骤详见有关Drive r的补充信息

说明: 1、参数号码后加“#“者为需要更改的,如02号参数实际应用中应设为 0; 2、控制方式含义 T:转矩控制 S:速度控制 P:位置控制 3、出厂设定为”*”者为出厂时未设置,需根据实际自行设置; 4、其余参数可使用出厂设置; 5、最后一栏为新Driver与之对应的参数(有些没有); 6、更详细的说明参见Driver手册(旧)或参考Driver手册(新)中对 应的参数; 7、以下表格为Driver实际参数设置表。

图解blast验证引物教程

图解blast验证引物教程 1、进入网页:https://www.wendangku.net/doc/8715200024.html,/BLAST/ 2、点击Search for short, nearly exact matches 3、在search栏中输入引物系列: 注:文献报道ABCG2的引物为5’-CTGAGATCCTGAGCCTTTGG-3’; 5’-TGCCCATCACAACATCATCT-3’ (1)输入方法可先输入上游引物,进行blast程序,同样方法在进行下游引物的blast程序。这种方法叫繁琐,而且在结果分析特异性时要看能与上游引物的匹配的系列,还要看与下游引物匹配的系列——之后看两者的交叉。 (2)简便的做法是同时输入上下游引物:有以下两种方法。输入上下游引物系列都从5’——3’。 A、输入上游引物空格输入下游引物

B、输入上游引物回车输入下游引物 4、在options for advanced blasting中: select from 栏通过菜单选择Homo sapiens【ORGN】Expect后面的数字改为10 5、在format中: select from 栏通过菜单选择Homo sapiens【ORGN】Expect后面的数字填上0 10

6、点击网页中最下面的“BLAST!” 7、出现新的网页,点击Format! 果。

(1)图形格式: 图中①代表这些序列与上游引物匹配、并与下游引物互补的得分值都位于40~50分 图中②代表这些序列与上游引物匹配的得分值位于40~50分,而与下游引物不互补 图中③代表这些序列与下游引物互补的得分值小于40分,而与上游引物不匹配 通过点击相应的bar可以得到匹配情况的详细信息。 (2)结果信息概要: 从左到右分别为: A、数据库系列的身份证:点击之后可以获得该序列的信息 B、系列的简单描述 C、高比值片段对(high-scoring segment pairs, HSP)的字符得分。按照得分的高低由大到小排列。得分的计算公式=匹配的碱基×2+0.1。举例:如果有20个碱基匹配,则其得分为40.1。 D、E值:代表被比对的两个序列不相关的可能性。【The E value decreases exponentially as the Score (S) that is assigned to a match between two sequences increases】。E值最低的最有意义,也就是说序列的相似性最大。设定的E值是我们限定的上限,E值太高的就不显示了 E、最后一栏有的有UEG的字样,其中: U代表:Unigene数据库 E代表:GEO profiles数据库 G代表:Gene数据库

Blast本地化详细流程

Blast 2.4.0+本地化详细流程(基于Windows系统) 1.程序获得。从NCBI上下载Blast本地化程序,下载地址: ftp://https://www.wendangku.net/doc/8715200024.html,/blast/executables/blast+/LATEST/ 64×安装版▲ 64×解压(绿色)版▲ 最好安装或解压到X盘根目录:如X:\blast,尽量简短,方便后边命令输入。 2.原始序列获得。方法1:找到转录组测序数据unigene数据库文件:unigene.fasta 或unigene.fa,若为unigene.fa则直接改后缀为.fasta即可。找到或修改后将数据库文件移动至Blast本地化程序目录“X:\blast\bin”。方法2:从NCBI中的ftp 库下载所需要库,链ftp://https://www.wendangku.net/doc/8715200024.html,/blast/db/FASTA/,其中nr.gz为非冗余的数据库,nt.gz为核酸数据库,month.nt.gz为最近一个月的核酸序列数据。下载的month.nt.gz先用WINRAR解压缩,然后用makeblastdb.exe格式化。方法3:利用新版blast自带的update_blastdb.pl进行下载,这需要安装perl程序。 注释:上述三种方法各有优缺点,前两种下载速度较快,但是每次进行检索都需要对数据库进行格式化(转化成二进制数据),第三种方法下载速度较慢,但是NCBI 中已经格式化好的,在进行本地检索时不需再进行格式化,直接用即可。 3.用文本编辑器(txt文件改名字及后缀)创建一个ncbi.ini文件,文件包含下 面内容:[NCBI]Data="C:\blast\data\" 先新建TXT文件,然后改属性,将ncbi.ini文件存放到C:\Windows 4.将Blast本地化程序目录添加路径中(该步骤非必须,但会给以后的操作带来 方便),方法: a)右击我的电脑选择属性,选择高级,点击环境变量,设置环境变量 b)系统变量中,选择Path,点击“编辑”,在变量值的后面添加Blast本地化 程序所在路径,E:\blast 点击确定,将安装路径添加到path。 5.运行MS-DOC。打开DOC窗口(点击开始,选择运行,打开的输入框中输 入“CMD”,确定),访问Blast本地化程序所在文件夹,依次输入:(1)X: 回车;(2)cd blast\bin,回车。

新力川伺服驱动使用说明

感谢您使用本产品,本使用操作手册提供LCDA系列伺服驱动器的相关信息。内容包括: ●伺服驱动器和伺服电机的安装与检查 ●伺服驱动器的组成说明 ●试运行操作的步骤 ●伺服驱动器的控制功能介绍与调整方法 ●所有参数说明 ●通讯协议说明 ●检测与保养 ●异常排除 ●应用例解说 本使用操作手册适合下列使用者参考: ●伺服系统设计者 ●安装或配线人员 ●试运行调机人员 ●维护或检查人员 在使用前,请您仔细详读本手册以确保使用上的正确。此外,请将它妥善保存在安全的地点以便随时查阅。下列在您尚未读完本手册时,务必遵守事项: ●安装的环境必须没有水气,腐蚀性气体或可燃性气体。 ●接线时,禁止将三相电源接至马达U、V、W的连接器,因为一旦接错 时将损坏伺服驱动器。 ●接地工程必须确实实施。 ●在通电时,请勿拆解驱动器、马达或更改配线。 ●在通电动作前,请确定紧急停机装置是否随时开启。 ●在通电动作时,请勿接触散热片,以免烫伤。 如果您在使用上仍有问题,请洽询经销商或者本公司客服中心。

安全注意事项 LCDA 系列为一开放型(Open Type )伺服驱动器,操作时须安装于遮蔽式的控制箱内。本驱动器利用精密的回授控制与结合高速运算能力的数字信号处理器(Digital Signal Processor,DSP ),控制IGBT 产生精确的电流输出,用来驱动三相永磁式同步交流伺服马达(PMSM )达到精准定位。 LCDA 系列可使用于工业应用场合上,且建议安装于使用手册中的配线(电)箱环境(驱动器、线材与电机都必须安装于符合环境等级的安装环境最低要求规格)。 在按收检验、安装、配线、操作、维护与检查时,应随时注意以下安全注意事项。 标志[危险]、[警告]与[禁止]代表的含义: ? 意指可能潜藏危险,若未遵守要求可能会对人员造成严 重伤或致命 ? 意指可能潜藏危险,若未遵守可能会对人员造成中度的 伤害,或导致产品严重损坏,甚至故障 ? 意指绝对禁止的行动,若未遵守可能会导致产品损坏, 或甚至故障而无法使用

本地blast的详细用法∷柳城

本地blast的详细用法 Posted on 03 四月 2009 by 柳城,阅读 9,626 本地blast的详细使用方法 blast all -p blastn -i myRNA.fasta -d humanRNA.fasta -o myresult.blastout -a 2 -F F -T T -e 1e-10 解释如下: blastall: 这是本地化/命令行执行blast时的程序名字!(Tips:blastall直接回车就会给出你所有的参数帮助,但是英文的) -p: p 是program的简写,program在计算机领域中是程序的意思。此参数是指定要使用何种子程序,所谓子程序,就是针对不同的需要,如核酸序列和核酸序列进行比对、蛋白质序列和蛋白质序列进行比对、假设翻译后核酸序列于蛋白质序列进行比对,选择相应的子程序: blastn 是用于核酸对核酸 blastp 是蛋白质对蛋白质序列等等,一共5个自程序。 -i: i 是input的简写,意思是输入文件,就是你自己的要进行比对的序列文件(fasta格式) -d: d是database的简写,意思是要比对的目标数据库,在例子中就是humanRNA.fasta (别忘了要formatdb) -o: o是output的简写,意思是结果文件名字,这个根据你自己的习惯起名字,可以带路径,(上边两个参数-i -d 也都可以带路径) *注意以上4个参数是必须的,缺一不可,下面的参数是为了得到更好的结果自己可调的参数,如果你不加也没有关系,blastall程序本身会给一个默认值! -a: 是指计算时要用的CPU个数,我的机器有两个CPU,所以用-a 2,这样可以并行化进行计算,提高速度,当然你的计算机就一个CPU,可以不用这个参数,系统默认值为1,就是一个CPU -F: 是filter的简写,blastall程序中有对简单的重复序列和低复杂度的一些repeats过滤调,默认是T (注意以后的有几种参数就两个选项,T/F T就是ture,真,你可以理解为打开该功能; F就是false,假,理解为关闭该功能) -T: 是HTML的简写,是指blast结果文件是否用HTML格式,默认是F!如果你想用IE看,我建议用-T T -e: 是Expectation value,期望值,默认是10,我用的10-10! BLASTALL 用法 a.格式化序列数据库 格式化序列数据库— —formatdb formatdb简单介绍: formatdb处理的都是格式为 ASN.1和FASTA,而且不论是核苷酸序列数据库,还是蛋白质序列数据库;不论是使用Blastall ,还是Blastpgp,Mega Blast应用程序,这一步都是不可少的。 formatdb命令行参数: formatdb - 得到formatdb 所有的参数显示(见附录二)和介绍, 主要参数的说明:

伺服驱动器参数设置方法

伺服驱动器参数设置方法 在自动化设备中,经常用到伺服电机,特别是位置控制,大部分品牌的伺服电机都有位置控制功能,通过控制器发出脉冲来控制伺服电机运行,脉冲数对应转的角度,脉冲频率对应速度(与电子齿轮设定有关),当一个新的系统,参数不能工作时,首先设定位置增益,确保电机无噪音情况下,尽量设大些,转动惯量比也非常重要,可通过自学习设定的数来参考,然后设定速度增益和速度积分时间,确保在低速运行时连续,位置精度受控即可。 1.位置比例增益:设定位置环调节器的比例增益。设置值越大,增益越高,刚度越大,相同频率指令脉冲条件下,位置滞后量越小。但数值太大可能会引起振荡或超调。参数数值由具体的伺服系统型号和负载情况确定。 2.位置前馈增益:设定位置环的前馈增益。设定值越大时,表示在任何频率的指令脉冲下,位置滞后量越小位置环的前馈增益大,控制系统的高速响应特性提高,但会使系统的位置不稳定,容易产生振荡。不需要很高的响应特性时,本参数通常设为0表示范围:0~100% 3.速度比例增益:设定速度调节器的比例增益。设置值越大,增益越高,刚度越大。参数数值根据具体的伺服驱动系统型号和负载值情况确定。一般情况下,负载惯量越大,设定值越大。在系统不产生振荡的条件下,尽量设定较大的值。 4.速度积分时间常数:设定速度调节器的积分时间常数。设置值越小,积分速度越快。参数数值根据具体的伺服驱动系统型号和负载情况确定。一般情况下,负载惯量越大,设定值越大。在系统不产生振荡的条件下,尽量设定较小的值。 5.速度反馈滤波因子:设定速度反馈低通滤波器特性。数值越大,截止频率越低,电机产生的噪音越小。如果负载惯量很大,可以适当减小设定值。数值太大,造成响应变慢,可能会引起振荡。数值越小,截止频率越高,速度反馈响应越快。如果需要较高的速度响应,可以适当减小设定值。 6.最大输出转矩设置:设置伺服驱动器的内部转矩限制值。设置值是额定转矩的百分比,任何时候,这个限制都有效定位完成范围设定位置控制方式下定位完成脉冲范围。本参数提供了位置控制方式下驱动器判断是否完成定位的依据,当位置偏差计数器内的剩余脉冲数小于或等于本参数设定值时,驱动器认为定位已完成,到位开关信号为 ON,否则为OFF。 在位置控制方式时,输出位置定位完成信号,加减速时间常数设置值是表示电机从0~2000r/min的加速时间或从2000~0r/min的减速时间。加减速特性是线性的到达速度范围设置到达速度在非位置控制方式下,如果伺服电机速度超过本设定值,则速度到达开关信号为ON,否则为OFF。在位置控制方式下,不用此参数。与旋转方向无关。7.手动调整增益参数 调整速度比例增益KVP值。当伺服系统安装完后,必须调整参数,使系统稳定旋转。首先调整速度比例增益KVP值.调整之前必须把积分增益KVI及微分增益KVD调整至零,然后将KVP值渐渐加大;同时观察伺服电机停止时足否产生振荡,并且以手动方式调整KVP参数,观察旋转速度是否明显忽快忽慢.KVP值加大到产生以上现象时,必须将KVP 值往回调小,使振荡消除、旋转速度稳定。此时的KVP值即初步确定的参数值。如有必要,经KⅥ和KVD调整后,可再作反复修正以达到理想值。 调整积分增益KⅥ值。将积分增益KVI值渐渐加大,使积分效应渐渐产生。由前述对积分控制的介绍可看出,KVP值配合积分效应增加到临界值后将产生振荡而不稳定,如同KVP值一样,将KVI值往回调小,使振荡消除、旋转速度稳定。此时的KVI值即初步确定的参数值。

NCBI在线BLAST使用方法与结果详解

N C B I在线B L A S T使用方法与结果详解 IMB standardization office【IMB 5AB- IMBK 08- IMB 2C】

N C B I在线B L A S T使用方法与结果详解 BLAST(BasicLocalAlignmentSearchTool)是一套在蛋白质数据库或DNA数据库中进行相似性比较的分析工具。BLAST程序能迅速与公开数据库进行相似性序列比较。BLAST结果中的得分是对一种对相似性的统计说明。 BLAST采用一种局部的算法获得两个序列中具有相似性的序列。 Blast中常用的程序介绍: 1、BLASTP是蛋白序列到蛋白库中的一种查询。库中存在的每条已知序列将逐一地同每条所查序列作一对一的序列比对。 2、BLASTX是核酸序列到蛋白库中的一种查询。先将核酸序列翻译成蛋白序列(一条核酸序列会被翻译成可能的六条蛋白),再对每一条作一对一的蛋白序列比对。 3、BLASTN是核酸序列到核酸库中的一种查询。库中存在的每条已知序列都将同所查序列作一对一地核酸序列比对。 4、TBLASTN是蛋白序列到核酸库中的一种查询。与BLASTX相反,它是将库中的核酸序列翻译成蛋白序列,再同所查序列作蛋白与蛋白的比对。 5、TBLASTX是核酸序列到核酸库中的一种查询。此种查询将库中的核酸序列和所查的核酸序列都翻译成蛋白(每条核酸序列会产生6条可能的蛋白序列),这样每次比对会产生36种比对阵列。 NCBI的在线BLAST: 下面是具体操作方法 1,进入在线BLAST界面,可以选择blast特定的物种(如人,小鼠,水稻等),也可以选择blast所有的核酸或蛋白序列。不同的blast程序上面已经有了介绍。这里以常用的核酸库作为例子。 2,粘贴fasta格式的序列。选择一个要比对的数据库。关于数据库的说明请看NCBI在线blast数据库的简要说明。一般的话参数默认。 3,blast参数的设置。注意显示的最大的结果数跟E值,E值是比较重要的。筛选的标准。最后会说明一下。 4,注意一下你输入的序列长度。注意一下比对的数据库的说明。 5,blast结果的图形显示。没啥好说的。 6,blast结果的描述区域。注意分值与E值。分值越大越靠前了,E值越小也是这样。7,blast结果的详细比对结果。注意比对到的序列长度。评价一个blast结果的标准主要有三项,E值(Expect),一致性(Identities),缺失或插入(Gaps)。加上长度的话,就有四个标准了。如图中显示,比对到的序列长度为1405,看Identities这一值,才匹配到1344bp,而输入的序列长度也是为1344bp(看上面的图),就说明比对到的序列要长一

blast验证引物教程1

图解blast验证引物教程 ——以文献报道的人类的ABCG2的引物为例 1、进入网页:https://www.wendangku.net/doc/8715200024.html,/BLAST/ 2、点击Basic BLAST中的nucleotide blast选项 3、完成2操作后就进入了Basic Local Alignment Search Tool界面 (1)在Enter Query Sequence栏中输入引物序列: 注:文献报道ABCG2的引物为5’-CTGAGATCCTGAGCCTTTGG-3’; 5’-TGCCCATCACAACATCATCT-3’ 简便的做法是同时输入上下游引物。输入上下游引物系列都从5’—3’。输入上游引物后,加上≥20个字母n,再输入下游引物,如下图:

(2)在Choose Search Set栏中: Database根据预操作基因的种属定了,本引物可选Human genomic + transcript或 Others (nr etc.)。本人倾向于选后者,觉得此库信息更多。如下图: (3)在Program Selection中:选择Somewhat similar sequences (blastn)项,如下图: (4)在此界面最下面:如下图 Show results in a new window项是显示界面的形式,可选可不选,在此我们选上了。关键要点击Algorithm parameters参数设置,进入参数设置界面。 4. 参数设置: (1)在General Parameters中:Expect thresshold期望阈值须改为1000,大于1000也可以; 在Word size的下拉框将数字改为7。如下图:

图解blast验证引物教程1

图解blast 验证引物教程 ——以文献报道的人类的ABCG2的引物为例 1、 进入网页:https://www.wendangku.net/doc/8715200024.html,/BLAST/ 2、 点击Basic BLAST 中的nucleotide blast 选项 3、 完成2操作后就进入了Basic Local Alignment Search Tool 界面 (1)在Enter Query Sequence 栏中输入引物序列: 注:文献报道ABCG2的引物为5’-CTGAGATCCTGAGCCTTTGG-3’; 5’-TGCCCATCACAACATCATCT-3’ 简便的做法是同时输入上下游引物。输入上下游引物系列都从5’— 3’。 输入上游引物后,加上≥20个字母n ,再输入下游引物,如下图: 生 物 秀

(2)在Choose Search Set 栏中: Database 根据预操作基因的种属定了,本引物可选Human genomic + transcript 或Others (nr etc.)。本人倾向于选后者,觉得此库信息更多。如下图: (3)在Program Selection 中:选择Somewhat similar sequences (blastn)项,如下图: (4)在此界面最下面:如下图 生物秀-专心做生物 w w w .b b i o o .c o m

Show results in a new window 项是显示界面的形式,可选可不选,在此我们选上了。关键要点击Algorithm parameters 参数设置,进入参数设置界面。 4. 参数设置: (1)在General Parameters 中:Expect thresshold 期望阈值须改为1000,大于1000也可以;在Word size 的下拉框将数字改为7。如下图: (2)Scoring Parameters 无须修改 (3)Filters and Masking 中,一般来说也没有必要改 5.点击最下面一栏的BLAST 按钮,如图: 6.点击BLAST 按钮后,跳转出现如下界面: 7. 等待若干秒之后,自动跳转出现显示BLAST 结果的网页。该网页用三种形式来显示blast 的结果。 生物秀-专心做生物 w w w .b b i o o .c o m

本地Blast

本地Blast使用说明 一、软件的下载安装 1.1安装流程 建议安装在非系统盘,如将下载的 BLAST 程序安装到 E:\blast,生成bin、doc 两个子目录,其中 bin 是程序目录,doc 是文档目录,这样就安装完毕了。 1.2 设置环境变量 右键点击“我的电脑”-“属性”,然后选择“高级系统设置”标签-“环境变量”(图1),在用户变量下方“Path”随安装过程已自动添加其变量值,即“E:\Blast\bin”。此时点击“新建”-变量名“BLASTDB”,变量值为“E:\Blast\db”(即数据库路径,图2)。 二、查看程序版本信息 点击 Windows 的“开始”菜单下的“运行”,输入“cmd”调出 MS-DOS 命令行,转到 Blast 安装目录,输入命令“blastn -version”即可查看版本,若能显示说明本地blast 已经安装成功。 三、使用 3.1本地数据库的构建 下载所需的数据(Fasta格式),将X 放到E:\blast\db 文件夹下,然后调出MS-DOS 命令行,转到E:\blast\db 文件夹下运行以下命令:格式化

数据库,命令为: makeblastdb -in 数据库文件 -dbtype 序列类型(核酸:nul;蛋白:prot)-title database_title-parse_seqids -out database_name-logfile File_Name 格式化数据库后,创建三个主要的文件——库索引(indices),序列(sequences)和头(headers)文件。生成的文件的扩展名分别是:.pin、.psq、.phr(对蛋白质序列)或.nin、.nsq、.nhr(对核酸序列)。而其他的序列识别符和索引则包含在.psi和.psd(或.nsi 和.nsd)中。 3.2核酸序列相似性搜索 blastn -db database_name -query input_file -out output_file -outfmt "7 qacc sacc qstart qend sstart send length bitscore evalue pident ppos" 备注:qacc:查询序列Acession号;sacc:目标序列Acession号; qstart qend:分别表示查询序列比对上的起始、终止位置; sstart send:分别表示目标序列比对上的起始、终止位置; length:长度; bitscore:得分; evalue:E-Value值; pident:一致性; ppos:相似性 3.3 查看并获取目标序列: blastdbcmd -db refseq_rna -entry 224071016 -out test.fa 可以从数据库中提取gi号为224071016的序列,并且以fasta格式存入文 件 3.4蛋白质序列相似性搜索 Blastp -db database_name-query input_file -out output_file -outfmt "7 qacc sacc qstart qend sstart send length bitscore evalue pident ppos" 3.5 查看并获取目标序列:重复3.3

核酸BLAST

核酸BLAST: ?blastn程式——核酸序列比对。 ?MegaBLAST——可搜寻一批EST序列、长序列cDNA或基因体序列。 BLAST——Basic Local Alignment Search Tool——核酸与蛋白质序列比对工具。BLAST网页提供BLAST(Basic Local Alignment Search Tool)程式、概述、使用说明与常见问题解答(网址:https://www.wendangku.net/doc/8715200024.html,/BLAST/)。 BLAST Program Selection Guide: https://www.wendangku.net/doc/8715200024.html,/blast/producttable.shtml#tab31

在做BLASTn的时候,系统会给出三个程序选项,分别是Highly similar sequences (megablast), More dissimilar sequences (discontiguous megablast),Somewhat similar sequences (blastn) 。 第一个选项megablast是对高度相似DNA序列间的比较。鉴别一段未知DNA序列的最好办法就是看看在公共数据库中这段序列是否存在。Megablast就是对那些具有高度相似(相似性95% 以上)的长序列片断所特别设计的一种序列比较工具。Megablast除了提供序列联配的显著性期望值域之外,还提供了一种百分值域。在进行序列比较时,用户可以同时调整这两个参数以优化搜索结果。 第二个选项discontiguous megablast,当序列之间的差异比megablast大时,一般选用这个程序。其算法的基本原理是将查询序列分为一个一个的小片断,我们把它叫做字,通过字与数据库序列相比较,如果能够精确匹配,则以这个字为种子向两边延伸,从而获得符合我们要求的相似性序列。discontiguous megablast所应用的字是不连续的,这使得他的搜索精确性在三种搜索程序中是最高的。其模板类型选项分为三种编码(0),非编码(1),两者都有(2)。在编码模式中,根据第三位碱基的摆动原理,只要第一个和第二个碱基能够精确匹配,那么第三个碱基可以忽略,不做比较。在字的长度相同的情况下,discontiguous megablast的精确度要高于blastn。 第三个选项Somewhat similar sequences (blastn),这个程序比较的序列其相似程度可以非常低。它采用的算法与discontiguous megablast相同,只不过它的字是连续的。Blastn的字要比megablast短,所以其精确度要高于megablast,但是运算速度要慢一些。 注:字是影响blast灵敏度的一个主要参数,其取值要根据具体情况具体而定。 NCBI BLASTn: https://www.wendangku.net/doc/8715200024.html,/public_documents/vibe/details/NcbiBlastn.html

NCBI中Blast种类及使用简介

NCBI中Blast种类及使用简介 NCBI中Blast种类简介 1. Blast Assembled Genomes 在一个选择的物种基因组序列中去搜索。 2.Basic Blast 2.1 nucleotide blast--- 用核酸序列到核酸数据库中进行搜索,包括3个程序 2.1.1 Blastn----核酸序列(n)到核酸序列数据库中搜索,是一种标准的搜索。 2.1.2 megablast----该程序使用“模糊算法”加快了比较速度,可以用于快速比较两大系列序列。可以用来搜索一匹ESTs序列和大的cDNA或基因组序列, 适用于由于测序或者其他原因形成的轻微的差别的序列之间的比较 2.1.3 discontiguous megablast----与megablast不同的是主要用来比较来自不同物种之间的相似性较低的分歧序列。 2.2 Protein Blast 2.2.1 Blastp ---蛋白质序列到蛋白质序列数据库中搜索,是一种标准的搜索。 2.2.2 psi-blast---位点特异迭代BLAST —用蛋白查询来搜索蛋白资料库的一个程式。所有被BLAST发现的统计有效的对齐被总和起来形成一个多次对齐,从这个对齐,一个位置特异的分值矩阵建立起来。这个矩阵被用来搜索资料库,以找到额外的显著对齐,这个过程可能被反复迭代一直到没有新的对齐可以被发现。 2.2.3 PHI-BLAST---以常规的表达模型为特别位置进行PSI - BLAST检索,找出和待查询序列具有一样的表达模型且具有同源性的蛋白质序列。 2.3 Translating BLAST 2.3.1 blastx----先将待查询的核酸序列按6 种读框翻译成蛋白质序列,然后将翻译出的蛋白质序列与NCBI 蛋白质序列数据库比较。 2.3.2 tblastn-----先将核酸序列数据库中的核酸序列按6 种读框翻译成

Blast本地化安装图解

Blast本地化:window平台下blast软件的安装boyun发表于 2009-07-09 17:08 | 阅读 1 views 1.对于windows 2000/xp 用户,下载blast- 2.2.18-ia32-win32.exe安装文件 ftp://https://www.wendangku.net/doc/8715200024.html,/blast/executables/LATEST/blast- 2.2.18-ia32-win32.exe 2.创建一个新目录,例如C:\blast,将下载的文件blast-2.2.18-ia32-win32.exe复制到该目录,双击这个文件,自解压产生bin、data、doc 三个目录,bin是程序目录,data是程序使用数据的目录,doc是文档目录。 表:bin目录中的程序 程序说明 bl2seq.exe进行两条序列比对 blastall.exe做普通的blast比对 blastclust.exe blastpgp.exe copymat.exe fastacmd.exe通过gi号,接收号等,在数据库中检索序 列 formatdb.exe格式化数据库 formatrpsdb.exe impala.exe makemat.exe megablast.exe megablast程序 rpsblast.exe seedtop.exe 3.用文本编辑器创建一个ncbi.ini文件,文件包含下面内容:[NCBI] Data="C:\blast\data\" 将ncbi.ini文件存放到系统的Windows 或者 WINNT目录。 4.将”C:\blast\bin”目录添加路径中(该步骤非必须,但会给以后的操作带来方便),方法:

NCBI在线BLAST使用方法与结果详解

N C B I在线B L A S T使用 方法与结果详解 This model paper was revised by the Standardization Office on December 10, 2020

N C B I在线B L A S T使用方法与结果详解 BLAST(Basic Local Alignment Search Tool)是一套在蛋白质数据库或DNA数据库中进行相似性比较的分析工具。BLAST程序能迅速与公开数据库进行相似性序列比较。BLAST结果中的得分是对一种对相似性的统计说明。 BLAST 采用一种局部的算法获得两个序列中具有相似性的序列。 Blast中常用的程序介绍: 1、BLASTP是蛋白序列到蛋白库中的一种查询。库中存在的每条已知序列将逐一地同每条所查序列作一对一的序列比对。 2、BLASTX是核酸序列到蛋白库中的一种查询。先将核酸序列翻译成蛋白序列(一条核酸序列会被翻译成可能的六条蛋白),再对每一条作一对一的蛋白序列比对。 3、BLASTN是核酸序列到核酸库中的一种查询。库中存在的每条已知序列都将同所查序列作一对一地核酸序列比对。 4、TBLASTN是蛋白序列到核酸库中的一种查询。与BLASTX相反,它是将库中的核酸序列翻译成蛋白序列,再同所查序列作蛋白与蛋白的比对。 5、TBLASTX是核酸序列到核酸库中的一种查询。此种查询将库中的核酸序列和所查的核酸序列都翻译成蛋白(每条核酸序列会产生6条可能的蛋白序列),这样每次比对会产生36种比对阵列。 NCBI的在线BLAST: 下面是具体操作方法 1,进入在线BLAST界面,可以选择blast特定的物种(如人,小鼠,水稻等),也可以选择blast所有的核酸或蛋白序列。不同的blast程序上面已经有了介绍。这里以常用的核酸库作为例子。 2,粘贴fasta格式的序列。选择一个要比对的数据库。关于数据库的说明请看NCBI在线blast数据库的简要说明。一般的话参数默认。 3,blast参数的设置。注意显示的最大的结果数跟E值,E值是比较重要的。筛选的标准。最后会说明一下。 4,注意一下你输入的序列长度。注意一下比对的数据库的说明。 5,blast结果的图形显示。没啥好说的。 6,blast结果的描述区域。注意分值与E值。分值越大越靠前了,E值越小也是这样。7,blast结果的详细比对结果。注意比对到的序列长度。评价一个blast结果的标准主要有三项,E值(Expect),一致性(Identities),缺失或插入(Gaps)。加上长度的话,就有四个标准了。如图中显示,比对到的序列长度为1405,看Identities这一值,才匹配到1344bp,而输入的序列长度也是为1344bp(看上面的图),就说明比对到的序

NCBI在线BLAST使用方法与结果详解

NCBI在线BLAST使用方法与结果详解 BLAST(Basic Local Alignment Search Tool)是一套在蛋白质数据库或DNA 数据库中进行相似性比较的分析工具。BLAST程序能迅速与公开数据库进行相似性序列比较。BLAST结果中的得分是对一种对相似性的统计说明。 BLAST 采用一种局部的算法获得两个序列中具有相似性的序列。 Blast中常用的程序介绍: 1、BLASTP是蛋白序列到蛋白库中的一种查询。库中存在的每条已知序列将逐一地同每条所查序列作一对一的序列比对。 2、BLASTX是核酸序列到蛋白库中的一种查询。先将核酸序列翻译成蛋白序列(一条核酸序列会被翻译成可能的六条蛋白),再对每一条作一对一的蛋白序列比对。 3、BLASTN是核酸序列到核酸库中的一种查询。库中存在的每条已知序列都将同所查序列作一对一地核酸序列比对。 4、TBLASTN是蛋白序列到核酸库中的一种查询。与BLASTX相反,它是将库中的核酸序列翻译成蛋白序列,再同所查序列作蛋白与蛋白的比对。 5、TBLASTX是核酸序列到核酸库中的一种查询。此种查询将库中的核酸序列和所查的核酸序列都翻译成蛋白(每条核酸序列会产生6条可能的蛋白序列),这样每次比对会产生36种比对阵列。 NCBI的在线BLAST:https://www.wendangku.net/doc/8715200024.html,/Blast.cgi 下面是具体操作方法 1,进入在线BLAST界面,可以选择blast特定的物种(如人,小鼠,水稻等),也可以选择blast所有的核酸或蛋白序列。不同的blast程序上面已经有了介绍。这里以常用的核酸库作为例子。

无刷直流电机驱动器说明书

无刷驱动器DBLS-02 一概述: 本控制驱动器为闭环速度型控制器,采用最近型IGBT和MOS功率器,利用直流无刷电机的霍尔信号进行倍频后进行闭环速度控制,控制环节设有PID速度调节器,系统控制稳定可靠,尤其是在低速下总能达到最大转矩,速度控制范围150~10000rpm。 二产品特征: 1、PID速度、电流双环调节器 2、高性能低价格 3、20KHZ斩波频率 4、电气刹车功能,使电机反应迅速 5、过载倍数大于2,在低速下转矩总能达到最大 6、具有过压、欠压、过流、过温、霍尔信号非法等故障报警功能 三电气指标 标准输入电压:24VDC~48VDC,最大电压不超过60VDC。最大输入过载保护电流:15A、30A两款连续输出电流:15A 加速时间常数出厂值:秒其他可定制 四端子接口说明: 1、电源输入端: 引角序号引角名中文定义 1V+直流+24~48VDC输入 2GND GND输入 引角序号引角名中文定义 1MA电机A相 2MB电机B相

3MC电机C相 4GND地线 5HA霍尔信号A相输入端 6HB霍尔信号B相输入端 7HC霍尔信号C相输入端 8+5V霍尔信号的电源线 G ND:信号地F/R:正、反转控制,接GND反转,不接正转,正反转切换时,应先关断EN E N:使能控制:EN接地,电机转(联机状态),EN不接,电机不转(脱机状态)B K:刹车控制:当不接地正常工作,当接地时,电机电气刹车,当负载惯量较大时,应采用脉宽信号方式,通过调整脉宽幅值来控制刹车效果。S V ADJ:外部速度衰减:可以衰减从0~100%,当外部速度指令接时,通过该电位器可以调速试机P G:电机速度脉冲输出:当极对数为P时,每转输出6P个脉冲(OC门输入)A LM:报警输出:当电路处于报警状态时,输出低电平(OC门输出)+5V:调速电压输出,可用电位器在SV和GND形成连续可调内置电位器:调节电机速度增益,可以从0~100%范围内调速。 五驱动器与无刷电机接线图 六机械安装:

血液循环驱动器概述及使用流程

气压式血液循环驱动器概述 (一)原理 气压式血液循环驱动器用于在进行间歇式气动压迫的过程中,充气压力带通过压迫肢体从而增强静脉血液的流动。在完一次压迫过程之后,主机将对静脉血管再次充满血液的时间进行测量,从而在经过相应时间段的等待之后,重新启动下一次压迫过程。从而有助于防止出现深静脉血栓和肺部栓塞(DVT/PE)。 (二)适应证 1、高风险手术全髋关节置换术,全膝关节置换术,髋关节骨折。 2、存在发生DVT风险的无禁忌证的患者。 3、可用于对抗凝治疗有禁忌的患者(如神经外科手术、头部创伤的患者等)。 (三)禁忌证 1.任何有可能妨碍充气压力带作用的腿局部情况,例如:皮炎,静脉结扎(在手术后即刻),坏疽或者刚做完皮肤移植手术。 2.严重的动脉硬化症或其他缺血性血管病。 3.腿部大范围水肿或由充血性心力衰竭引发的肺部水肿。 4.腿部严重畸形。 5.疑似已出现深静脉血栓。 (四)尺寸的选择 腿长<55. 9cm,选择小号充气压力带。 腿长在55. 9~71. 7cm之间,选择中号充气压力带。 腿长>71. 1 cm,选择大号充气压力带。 (五)人员资格 1.具有执业资格证书的护士。 2.经过“气压式血液循环驱动器护理指南”培训合格的护士。 (六)养护要点 1.使用时动作轻柔,保证管路畅通。 2.及时收回,放置固定地点保存。 3.将充气压力带、连接管整理好,管路勿打折保存。 4.发现污垢及时清理干净。 5.如使用过程中发现异常及时送专业维修部门进行维修。 (七)评估要点 1.评估患者患肢伤口情况、精神状况及配合程度。 2.评估患肢是否有深静脉血栓。 3.评估患者皮肤是否有破损。 (八)宣教要点 1.告知患者使用气压式血液循环驱动器的作用和意义。 2.患者治疗过程中如有不适,及早通知护士。

直流无刷电机驱动器说明书(1)

BLDC无刷电机驱动器 (UB510) 使用手册w w w.u p u ru.c o m

感谢您使用本产品,本使用操作手册提供UB510驱动器的配置、调试、控制相关信息。内容包括。 l驱动器和电机的安装与检查 l试转操作步骤 l驱动器控制功能介绍及调整方法 l检测与保养 l异常排除 本使用操作手册适合下列使用者参考 l安装或配线人员 l试转调机人员 l维护或检查人员 在使用之前,请您仔细详读本手册以确保使用上的正确。此外,请将它妥善放置在安全的地点以便随时查阅。下列在您尚未读完本手册时,请务必遵守事项: l安装的环境必须没有水气,腐蚀性气体及可燃性气体 l接线时禁止将电源接至电机 U、V、W 的接头,一旦接错时将损坏驱动器 l在通电时,请勿拆解驱动器、电机或更改配线 l在通电运作前,请确定紧急停机装置是否随时启动 l在通电运作时,请勿接触散热片,以免烫伤 警告: 驱动器用于通用工业设备。要注意下列事项: (1).为了确保正确操作,在安装、接线和操作之前必须通读操作说明书。 (2).勿改造产品。 (3).当在下列情况下使用本产品时,应该采取有关操作、维护和管理的相关措施。在这种情况下,请与我们联系。 ①用于与生命相关的医疗器械。 ②用于可能造成人身安全的设备,例如:火车或升降机。 ③用于可能造成社会影响的计算机系统 ④用于有关对人身安全或对公共设施有影响的其他设备。 (4).对用于易受震动的环境,例如:交通工具上操作,请咨询我们。 (5).如未按上述要求操作,造成直接或间接损失,我司将不承担相关责任。

1概述 本公司研发生产的BLDC驱动器是一款高性能,多功能,低成本的带霍尔传感器直流无刷驱动器。全数字式设计使其拥有灵活多样的输入控制方式,极高的调速比,低噪声,完善的软硬件保护功能,驱动器可通过串口通信接口与计算机相连,实现PID参数调整,保护参数,电机参数,加减速时间等参数的设置,还可进行IO输入状态,模拟量输入,告警状态及母线电压的监视。 1.1驱动器参数列表 输入电压DC18V-50V 工作电流<=10A 电机霍尔类型60度,300度,120度,240度 工作模式霍尔速度闭环 调速方式0-5V模拟量输入, 0-100%PWM输入(PWM频率范围:1KHz-20KHz) 内部给定, 多段速1, 多段速2, 调速范围0—6000RPM 保护功能l短路:当异常电流大于50A时,产生短路保护 l过流:当电流超过工作电流设置值并持续一设定时间 后产生过流保护 l过压:当电压超过55V时产生过电压保护。 l欠压:当电压低于18V时产生欠电压保护。 l霍尔异常:包括相位异常及值异常. 工作环境场合:无腐蚀性,易燃,易爆,导电的气体,液体,粉尘 温度:-10-55 ℃(无冻霜) 湿度:小于90%RH(不结露) 海拔:小于1000m 振动:小于0.5G, 10hz—60hz(非连续运行) 防护等级:IP21 散热方式自然风冷 尺寸大小120*76*33 单位:mm 重量250g 1.2特点 l速度PID闭环控制,低速转矩大, l调速范转宽,0-6000RPM l运行加减速时间可由软件设定,实现平滑柔和运行。 l驱动器自身损耗小,效率高,温升低,因此体积小,易安装 l多种速度控制方式,由软件设定。 l使能,方向,刹车输入信号的极性可由软件设定 l多种完善的保护功能。 l内置刹车电阻及控制电路(可选),用于消耗再生能量,防止过电压。 2接口定义与连接图

相关文档