Đọc tin (hay) khoa học hộ bạn!!!

**haichit** · 11-03-2010

Lâu rồi không post bài nào, tớ mở cái topic này ra để ace post những thông tin khoa học mà mình thích. Vì đây là ở góc chia sẻ tài nguyên nghiêm túc nên mọi người post bài từ những nguồn nghiêm túc chứ không phải báo lá cải kiểu dantri, vnexpress nhé. Và nên có vài dòng description để trả lời câu hỏi 'tại sao ngươi khác cần đọc tin này'.

**haichit** · #2 (**permalink**) 11-03-2010

Có lẽ DE Shaw sẽ là một trong những người tạo ra evolution cho computational biology. Hiện tại thì các simulation cho protein đều giới hạn ở nanosecond time scale, nghĩa là còn lâu mới được tới microsecond và xa hơn là second time scale. Đây có lẽ là bước tiến khá quan trọng.

	Trích:
	Scientists have long been frustrated in their efforts to use computers to simulate the atomic detail of how proteins fold into their three-dimensional structures. The computing demands for simulating all the motions of a protein's atoms and the surrounding water are so high that scientists have had difficulty tracking the myriad atomic wiggles and gyrations for long enough to see the complete folding process. But now help is on the way. On page [Only registered and activated users can see links. ], computational biologists led by computer scientist and former hedge fund manager David Shaw report that they ran a specially built supercomputer for about 3 weeks to simulate a relatively small protein going through 15 rounds of folding and unfolding over 200 microseconds. They also tracked the folding gyrations of a similarly sized protein for more than a millisecond.

[Only registered and activated users can see links. ]

Mê Nấm · #3 (**permalink**) 11-03-2010

Cám ơn bạn chit. Không biết cái này có giúp gì về dự đoán cấu trúc từ trình tự không nhỉ?

**haichit** · #4 (**permalink**) 11-03-2010

Nếu bác hỏi câu này 10 năm trước thì chưa có câu trả lời nhưng mà bây giờ thì có rồi, phần lớn những small peptide đều được dự đoán cấu trúc từ sequence cả. Có nhiều helix và hairpin từ vài đến vài chục residue đều có thể dự đoán được cấu trúc 3D và có RMSD tầm 1.2 Angstrom so với cấu trúc từ NMR hoặc X-ray luôn.

Nếu bác thích thì có thể đọc 1 paper từ lab của haichit về cái này.
[Only registered and activated users can see links. ]

	Trích:
	All-Atom Structure Prediction and Folding Simulations of a Stable Protein Carlos Simmerling,, , Bentley Strockbine, and, Adrian E. Roitberg, Journal of the American Chemical Society 2002 124 (38), 11258-11259

Tất nhiên là mấy cái prediction này chỉ dừng lại ở small protein thôi chứ chưa áp dụng được ở những big protein tầm vài trăm residues.

**haichit** · #5 (**permalink**) 11-04-2010

GPU Test Drive — Get 10x Faster Molecular Dynamics

	Trích:
	Signup for a FREE trial of Tesla GPUs today and experience the acceleration that other scientists are publishing with. Simply register online to receive an email with log-in instructions to simulate your model in a workstation powered by Tesla GPUs.

[Only registered and activated users can see links. ]

Mê Nấm · #6 (**permalink**) 11-05-2010

Hihi. Khi nào có supercomputer simulation được large protein folding hẳn vài chục giây, và có model hợp lý, thì dẹp được X-ray với NMR, và các bác làm protein structure khỏi có cơ hội lên Nature như đi chợ.

:
Còn đã dự đoán, thì vẫn là đoán mò, và protein structure prediction vẫn là holy grail của ngành bioinformatics.

**SpringerCV** · #7 (**permalink**) 11-18-2010

From Superpower to Supercomputer
For the first time ever, a Chinese system tops the list of the fastest supercomputers in the world
--------------------------------------------------

On Sunday night, the Supercomputing 10 conference, in New Orleans, released a new Top 500 list of the fastest computers on the planet. China tops the list for the first time with the Tianhe-1A system, at the National Supercomputer Center in Tianjin, clocking in at 2.57 petaflops per second. The list helps highlight the latest trends in computing—the power of general microprocessors, the ascendance of Asian computing, and the importance of interconnect technologies—Tianhe-1A uses a novel interconnect technology believed to be twice as fast as InfiniBand. Host Steven Cherry talks with Jack Dongarra of the University of Tennessee and the Oak Ridge National Laboratory, whose Jaguar supercomputer ranks second on the new list.

[Only registered and activated users can see links. ]

Mê Nấm · #8 (**permalink**) 11-29-2010

Có khả năng đã tìm thấy liều thuốc trường sinh bất lão cho con người.
[Only registered and activated users can see links. ]
Bài báo gốc:
[Only registered and activated users can see links. ]

Về sự lão hóa bình thường của con người, một trong những nguyên nhân chính là do nhiễm sắc thể, cấu trúc mang thông tin di truyền của con người, bị ngắn lại một tí sau mỗi lần thế bào phân chia (để thay mới). Khi nhiễm sắc thế ngắn đến một mức nhất định, tế bào sẽ chết và cơ thể sẽ lão hóa. Ở những tế bào bất tử, enzyme telomerase hoạt động để duy trì nhiễm sắc thể ở một trạng thái không quá ngắn để gây chết với tế bào. Bình thường tất cả các tế bào trong cơ thể con người đều có tiềm năng bất tử, bằng cách hoạt hóa enzyme telomerase này. Nhưng vì số phận, một số tế bào trở nên bất tử, còn phần lớn các tế bào còn lại đều lão hóa và chết. Nếu có thể tìm ra loại thuốc hoạt hóa telomerase ở các tế bào không bất tử, ta sẽ đảo ngược được quá trình lão hóa. Tất nhiên nếu lạm dụng thuốc này quá, các tế bào này sẽ trẻ mãi không già, đẻ sòn sòn năm bảy lứa, và kết cục là...ung thư

**haichit** · #9 (**permalink**) 01-11-2011

Trends in computational biology—2010

Nature Biotechnology phỏng vấn 15 nhà khoa học hàng đầu trong ngành này.

PS: Những ai học CS mà đá ngang qua computational biology thì đúng làm làm mưa làm gió, dễ dàng bóp mũi mấy người học Bio, hoặc Chem.

Article: The field of computational biology encompasses a set of investigative tools as much as being a research endeavor in its own right. It is often difficult to gauge the utility and significance of a computational tool, at least until the research community has had sufficient time to explore, exploit and hone it in various applications. In an effort to identify recent notable breakthroughs in the field of computational biology, Nature Biotechnology surveyed leading researchers in the area, asking them to nominate papers of particular interest published in the previous year that have influenced the direction of their research. Some of the nominated papers had been featured in our pages and elsewhere; others were completely off our radar. Although we surveyed a small group of 15 scientists, the nominated papers (Box 1) provide a snapshot of some of the most exciting areas of current computational biology research.

[Only registered and activated users can see links. ]

**haichit** · #10 (**permalink**) 02-03-2011

[Only registered and activated users can see links. ]

Article: Cloud Computing—What's in It for Me as a Scientist?
Armando Fox

+ Author Affiliations
Department of Computer Science, University of California, Berkeley, CA 94720, USA.
E-mail: [Only registered and activated users can see links. ]

Many scientists would love access to large-scale computational resources but find that the programming demands of using a supercomputer—as well as the cost and queuing time—are too daunting. Privately owned cloud computers—large data centers filled with computers that mainly run their company's software—are now becoming available to outside users, including scientists and educators. Companies are leasing their computing resources on demand from a large shared pool to individuals who run their own software on a pay-as-you-go basis. This approach is an example of cost associativity (1): 1000 computers used for 1 hour costs the same as one computer used for 1000 hours. If your problem can be computed in a way that takes advantage of parallel processing, you can now get the answer 1000 times as fast for the same amount of money.

[Only registered and activated users can see links. ]
This view of a Microsoft data center is a tiny fraction of the servers available for cloud computing.

CREDIT: MICROSOFT CORPORATION

Although companies had long been operating “private clouds” that run programs such as Google Search or Microsoft Hotmail, Amazon was the first to let outside users run software on their computers. For example, Amazon's Elastic Compute Cloud (EC2), announced in late 2007, allows anyone with a credit card to use any number of computers in Amazon's data centers for 8.5 cents per computer-hour with no minimum or maximum purchase and no contract. Such an arrangement is possible because these “warehouse-scale” data centers (∼50,000 servers, see the figure) are five to seven times cheaper to build and operate than smaller facilities (∼1000 servers) (2), in terms of network, storage, and administrator costs. When operational costs, which include human administrators, power, and networking, are considered, the charges for cloud computing to outside users is price-competitive with using in-house facilities.

Cost-associativity enables new capabilities. For example, a researcher in our laboratory created an automated classifier to detect spam on the popular social-communication site Twitter.com. Training the classifier took about 270 hours on a typical desktop workstation. The training program could be parallelized: The problem could be broken into pieces and run at the same time, only occasionally sharing results between different pieces. The same task took 3 hours using about 100 servers in Amazon's cloud (about $250 in usage fees). Many universities have also begun to use cloud computing for education, where cost-associativity is a great fit for semester courses. Lots of computing demand could be provided around assignment deadlines (more than even the biggest schools could provide), and when demand is less (between deadlines), no outside resources need to be purchased.

Initially, cloud-computing hardware was configured primarily for its earliest adopters—Web-based applications—and early attempts to run scientific applications on the cloud gave discouraging results (3, 4). New hardware is now configured for better performance on scientific applications. For example, Amazon's recently added “cluster computing instances,” priced at $1.60 per computer-hour, run scientific benchmarks 8.5 times as fast as the original cloud hardware, according to experiments at the National Energy Research Scientific Computing Laboratory at Lawrence Berkeley National Laboratory (5).

Cloud computing works best when a problem can be broken down into a large number of relatively independent tasks, each running on its own computer. Software frameworks like Google's MapReduce (6) and its open-source equivalent Hadoop (7) provide a data-parallel “building block” for expressing such computations (much like a Web design framework allows you to “build” a Web site by filling in the relevant information and functions you want). Critically, these frameworks also hide the complex software machinery that handles inevitable transient machine failures when hundreds of machines in a cloud environment work on a problem simultaneously. Many of the “success stories” of science in the cloud have embraced Hadoop, and other popular tools such as the statistical package R now feature libraries that integrate with it. However, many problems cannot be easily expressed in terms of map and reduce tasks (mapping parcels out the work, and reduce collates the results). Even when they can, the programming effort required may be substantial.

Most desktop software is not written to take advantage of cloud computing and requires modification before it could harness cloud resources and run faster. However, the popular packages MATLAB and Mathematica are now available in versions that can “farm out” work to a public cloud. Cloud vendors including Amazon and IBM are working with independent software vendors on cloud-friendly versions of popular scientific software.

What about software designed for parallel supercomputers? Programs written to the MPI (message-passing interface) standard run in the cloud, but the performance of such programs is sensitive to whether all the machines run in tight lockstep when working a problem (a characteristic of supercomputers but one that is not necessarily part of the cloud-computing architecture). Still, cloud technology suppliers such as Intel, Advanced Micro Devices, and VMware are adding hardware and software features necessary to improve MPI performance in the cloud, a move that highlights new “buying power” in the scientific computing community.

Many scientific computing problems do not require supercomputer performance but would benefit greatly from modest parallelism—say, tens or hundreds of computers in the cloud. The total “time to answer” may still be quicker, according to Foster (8), even for larger problems because a cloud-based supercomputer can be provisioned and running in minutes rather than hours or days, without waiting in a queue.

Some problems are so data intensive that they demand large-scale computing. The Large Synoptic Survey Telescope in Chile may generate up to 30 terabytes of data daily. At typical long-haul network speeds of around 20 megabits per second (9), 30 terabytes would take more than 4 months and $3000 in network charges to copy to Amazon's cloud. Many cloud providers now allow users to ship crates of hard drives to be physically incorporated into the cloud infrastructure, an idea championed by the late Jim Gray.

The lure of improved performance has already drawn scientists and engineers to use cloud computing (10) in research on a number of topics, including large-population genetic risk analysis, information retrieval, and particle physics. Cloud computing installations are also being established by academic-industrial consortia [see, for example, (11–13)] to further encourage adoption of cloud computing by scientists.