Saturday, February 20, 2010

Making hay out of power law distributions

Gregory Mone at AOLNews discusses a paper published in the 19 Feb 2010 issue of Science by Albert-László Barabási and co-workers about analysis of data from 50,000 anonymous cell phone users to make predictions about human mobility, to show "where we are and when we are there" is quite foreseeable.

The evolution of Barabasi's publications in Science and Nature is of interest.

Barbasi's first paper on network science was submitted (while he was a postdoc at IBM) to four journals, including Science and Nature. All four rejected it. He remembers, “Nobody said it was wrong. But somehow the tone I got from the referee reports was, ‘Who cares?’”

Barbasi went to Notre Dame and worked in the area of materials science. One of his doctoral students, Réka Albert, co-authored a paper titled “What Keeps Sandcastles Standing?” that made the cover of Nature.

Science published a paper on the power-law distribution of links to nodes in the
Notre Dame internet network in October 1999. And he was off and running.

See Looking for the next big thing, which includes the text: The links in appeared to be distributed
in a much different pattern — a power law. A graph of a power law resembles a hockey stick, with a single steep curve sweeping downward into a long, long tail. There’s
nothing bell curvish about it.
and the text Internet Movie Database, a massive record of thousands of actors linked to each other by roles in the same films. (It’s the basis of the game “Six Degrees of Kevin Bacon.”)

IPBiz notes that that citations to journals are very much dependent on the identity of the journal. When one hears “20 percent of people hold 80 percent of wealth,” which describes a power law, one notes that only a few journals are responsible for the bulk of scientific citations. This disparity led to the creation of the concept of the "impact factor." Within this discussion, one found that WITHIN one journal a select group of papers are primarily responsible for the impact factor [about 90% of Nature's 2004 impact factor was based on only a quarter of its publications ]
See also Eugene Garfield, The History and Meaning of the Journal Impact Factor which includes the text: "The so-called 80/20 phenomenon applies, in that 20% of articles may account for 80% of the citations."

Note also the published US patent application DISEASE DIAGNOSES-BASES DISEASE PREDICTION, 20080183454, with first claim:

A system for predicting future disease for a subject comprising:a population information set comprising population disease diagnoses for members of a population;a subject-specific information set comprising at least one subject-specific disease diagnosis; anda diagnoses-based prediction module configured to predict one or more future diseases for the subject based on said subject-specific disease diagnosis and said population disease diagnoses for population members having at least one disease in common with the subject.

****Of relevance to metabolomics, from the article by Dale Keiger:

Next, Barabási teamed up with a Hungarian friend at Northwestern University, the
biologist Zoltán Oltvai, to examine metabolic networks —
genes linked by chemical reactions — in 43 different living
organisms. In every case they found the same power-law distribution
of nodes and links. The Internet, actors in Hollywood,
the proteins that regulate cellular metabolism, all
could be studied as networks with a common structure and
common mathematical laws.

For example: Vazquez Alexei; de Menezes Marcio A; Barabási Albert-László; Oltvai Zoltan N --
Impact of limited solvent capacity on metabolic rate, enzyme activities, and metabolite concentrations of S. cerevisiae glycolysis.


Post a Comment

<< Home