<!------------Discussion----------->
At present, approximately half of all human proteins have a recognizable domain 
using SuperFamily. However if no relevant crystallographic or NMR 
structure has been determined, no match will appear in this track even though a 
well-known domain might readily be identifiable by alignment methods. 
Domains of membrane proteins and difficult-to-study proteins are therefore
under-represented in the track, leading to a bias in favor of the smaller 
soluble proteins found at the Protein Data Bank 
(<A HREF="http://www.rcsb.org/" TARGET=_blank>PDB</A>).
Furthermore, some proteins have been studied 
as truncated, more-easily crystallized fragments and not all domains will 
have been detected.  These biases are seen clearly in the 
<A HREF="http://scop.mrc-lmb.cam.ac.uk/scop/count.html" 
TARGET=_blank>statistical breakdown</A> 
of the 1073 currently known superfamilies (all species, including 733 in human). 
<P>
Superfamilies go so far back in evolutionary time that branches have 
sometimes evolved quite different functions despite retaining the same fold. 
These functions may be distinct but plausibly related, as in the alkaline 
phosphatase superfamily where 17 human members have become 
<A HREF="http://www.mad-cow.org/00/annotation_frames/tools/genbrow/sulfatases/sulfatases.html" 	
TARGET=_blank>sulfatases</A>. Or, the functions may be so diverged, as in the 
<A
HREF="http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6WK7-46M56WD-R&_coverDate=08%2F30%2F2002&_alid=145332535&_rdoc=1&_fmt=&_orig=search&_qd=1&_cdi=6899&_sort=d&view=c&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=1fc1573b1825d89fb59db24b9b88d87e"
TARGET=_blank>TIM beta/alpha-barrel family</A>, that 
multiple independent evolutionary development of the same fold
seems possible. Thus, functional interpretation of a 
match in the domain track requires a measure of caution. 
<P>
For humans, the current 
<A HREF="http://supfam.mrc-lmb.cam.ac.uk/SUPERFAMILY/cgi-bin/gen_list.cgi?genome=hs"
TARGET=_blank>distribution of observed domains</A> is dominated 
by thousands each of zinc fingers, 
immunoglobulin domains, P-loop NTPases, laminins, cadherins, olfactory genes, 
fibronectins, and protein kinases. At the other extreme, 114 domains are known 
only from single protein representatives. 
<P>
Some functional classes, such as extra-cellular matrix proteins, commonly 
have both multiple domain types and multiple copies of a given domain. 
Human proteins 
average only 1.5 currently detectable domains per protein and a large number
of superfamilies have but one domain (although these are mostly in small families). Mammalian 
proteins are some 50% larger than the average <em>E. coli</em> protein (450 vs 300 amino 
acids) but some of this is attributable to ragged growth at poorly conserved N- 
and C- termini. 

