1. Protein Bioinformatics Course
Matthew Betts & Rob Russell
AG Russell (Protein Evolution)
Course overview
Day 1 - Modularity
Day 2 - Interactions
Day 3 - Modularity & Interactions
Day 4 - Structure
Day 5 - Structure & Interactions
Daily schedule
10:00-11:00 lecture
11:00-12:00 work on exercises in pairs
12:00-13:00 lunch
13:00-15:30 work on exercises in pairs
16:00-17:00 presentations by you

2. Protein Sequence Databases

3. Database Searching Homologues = proteins with a common ancestor
Homology --> similar function
Sequence similarity --> homology
Find homologues using:
BLAST
Profile Searching

5. Scores and E-values
How much would I expect to get >= this score by chance alone?
How similar is my sequence to one in the database?
cf. random sequences
E = 1: one such match by chance
E < 0.01: significant
Depends on database:
size: larger = better
composition (random assumed)
Alignment
Substitution matrix
Gap penalties

6. Homology comes in two main types: Orthology and Paralogy
What is the difference and why does this matter?

7. Paralogues Paralogues Duplication - Speciation - - Speciation
Orthologues
Speciation -
- Speciation
Paralogues
Duplication -

8. Different Fates Orthologues:
Both copies required (one in each species)
conservation of function (‘same gene’)
adaptation to new environment
Easier to transfer
knowledge of function
between orthologues
Paralogues:
Both copies useful
conservation of function
One copy freed from selection
disabled
new function
Different parts of each free from selection
function split between them

9. Assignment of orthology / paralogy can be complicated by:
duplication preceding speciation
lineage-specific deletions of paralogs
complete genome duplications
many-to-one relationship
multi-domain proteins

10. Homology usually found by sequence similarity, but
…proteins with dissimilar sequences can still be homologous
Betts, Guigo, Agarwal, Russell, EMBO J 2001

11. Proteins are modular
Since the early 1970s it has been observed that protein structures are divided into discrete elements or domains that appear to fold, function and evolve independently.

12. Given a sequence, what should you look for?
Functional domains (Pfam, SMART, COGS, CDD, etc.)
Intrinsic features
Signal peptide, transit peptides (signalP)
Transmembrane segments (TMpred, etc)
Coiled-coils (coils server)
Low complexity regions, disorder (e.g. SEG, disembl)
Hints about structure?

13. Given a sequence, what should you look for?
“Low sequence complexity”
(Linker regions? Flexible? Junk?
Transmembrane segment
(crosses the membrane)
Signal peptide
(secreted or membrane
attached)
Tyrosine kinase
(phosphorylates Tyr)
Immunoglobulin domains
(bind ligands?)
SMART domain ‘bubblegram’ for human fibroblast growth factor (FGF) receptor 1
(type P11362 into web site: smart.embl.de)

14. Protein Modularity discrete structural and functional units
found in different combinations in different proteins
Receptor-related tyrosine-kinase
Non-receptor tyrosine-kinases
consider separately in predictions

15. Finding Protein Domains
through partial matches to whole sequences:
compare to databases of domains (Pfam, SMART, Interpro)
can be separated by:
low-complexity and disordered regions (SEG)
trans-membrane regions (TMAP)
coiled-coils (COILS)
query sequence:
match
Repeat searches using each domain separately

16. 12 000 domain alignments make sequence searching easier
WPP domain alignment
Alignments provide more information about a protein family and thus allow for more sensitive sequences than a single sequence.
Domain alignments also lack low-complexity or disorder (normally) and other domains that can make single sequence searches confusing.

17. Finding domains in a sequence

18. at the border of sequence detectability
Cryptic domains:
at the border of sequence detectability
Identified using more sensitive fold recognition methods that use structure to help find weak members of sequence families.
If Pfam or SMART or similar do not find a domain, and the region is probably not disordered, then fold recognition might help.
Gallego et al, Mol Sys Biol 2010

19. Domain peptide interactions
Recognition of ligands or targeting signals
Post-translational modifications

20. Linear motifs
Peptides interacting with a common domain often show a common pattern or motif usually 3-8 aas.
3BP1_MOUSE/ APTMPPPLPP
PTN8_MOUSE/ IPPPLPERTP
SOS1_HUMAN/ VPPPVPPRRR
NCF1_HUMAN/ SKPQPAVPPRPSA
PEXE_YEAST/ MPPTLPHRDW
SH3-interacting motif PxxP
“instance”
“motif”
“perpetrator”
“victim”
Puntervol et al, NAR, 2003; (Eukaryotic Linear Motif DB)

21. Linear motifs versus domains
Domains: large globular segments of the proteome that fold into discrete structures and belong in sequence families.
Linear motifs: small, non-globular segments that do not adopt a regular structure, and aren’t homologous to each other in the way domains are.
Motifs lie in the disordered part of the proteome.

22. Intrinsically unstructured or disordered proteins or protein fragments

23. (IUPred, RONN, DisORPred, etc)
Disorder predictors
(IUPred, RONN, DisORPred, etc)

24. Linear motif mediated interactions
2424
Linear motif mediated interactions
are everywhere
Include motifs for:
Targeting – e.g. KDEL
Modifications – e.g. phosphorylation
Signaling – e.g. SH3
About 200 are currently
known, likely many more
still to be discovered
Neduva & Russell, Curr. Opin. Biotech, 2006

25. Finding linear motifs in a sequence
Linear motifs are much harder to find than domains.
Long (>30 AA), belong to sequence families that help detect new family members
Short (typically < 8AA), simple patterns, e.g. PxxP will occur in most sequences randomly.