1. Sequence alignment & Substitution matrices By Thomas Nordahl

2. Sequence alignment
Sequence alignment is the most important technique used in bioinformatics
Infer properties from one protein to another
Homologous sequences often have similar biological functions
Most information can be deduced from a sequence if the 3D-structure is known
3D-structure determination is very time consuming (X-ray, NMR)
Several mg of pure protein is required (> 100mg)
Make crystal, solve structure, 1-3 years
Large facilities are needed to produce X-ray
Rotating anode or synchrotron
Determining primary sequence is fast, cheap
Structure more conserved than sequence

3. Growth of GenBank and WGS

4. Structures in PDB
Genbank

5. Car parts – analogy to protein folds
A fold: major structural similarity

6. Protein class & folds
A fold: major structural similarity

7. Structures in SCOP database
A fold: major structural similarity
The “world” seems to
consist of approx1400
protein folds.
Until 2014 no new folds
have been observed

8. What can we learn from sequence alignment
Find similar sequence from another organism
Information from the known sequence can be inherited
Layers of conserved information:
Structure > function > sequence where, ‘>’ means more conserved than
Structure (3D) is the most conserved feature
Proteins with different function may still share the same structure
Proteins with different may still share the same function
Often same function if 40-50% sequence identity
Often same protein fold if above 30% sequence identity
A fold: major structural similarity

9. Sequence alignment M V S T A 1
M V S T A M A T S A Antal identiske aa, % id ? Alignment score using identity matrix? Similar amino acids can be substituted, therefore other types of substitution matrices are used.

10. Blosum matrices
Blosum matrices are the most commonly used substitution matrices - Blosum50, Blosum62, blosum80
Symmetrical 20 x 20 matrix,
where each element is the
substitution score.
Positive scores:
Amino acids are likely to be
aligned in a sequence alignment
They share similar chemical
characteristics
Negative scores:
Less likely substitution –
but still occur.
Zero Scores:
Invariant
Q) In an alignment what is the most likely amino acid that Arg will align to
besides itself?

11. Log-odds scores Log-odds scores are given by
Log( Observation/Expected)
The log-odd score of matching amino acid j with amino acid i in an alignment is
where Pij is the frequency of observation i aligned with j, and Qi, Qj are the frequency if amino acids i and j in the data set.
The log-odd score is (in bit units)
Where, Log2(x)=logn(x)/logn(2)
S has been normalized to half bits, therefore the factor 2

12. Example of a scoring matrix BLOSUM80
A R N D C Q E G H I L K M F P S T W Y V A R N D C Q E G H I L K M F P S T W Y V
Log-Odds scores
Have been rounded off
to integers

13. An example NAA = 14 1 2 3 4 seq1: V V A D seq2: A A A D seq3: D V A D
Sij = 2log2(Pij/(QiQj))
Pij can be calculated as Nij/(Sumij Nij), where
Nij is the number of times amino acid i is aligned to amino acid j
Sum Nij is the total number of all alignments Nij
Qi is the frequency observed in alignment of amino acid i
MSA – Multiple Sequemce Alignment
How to calculate NAA
seq1: V V A D
seq2: A A A D
seq3: D V A D
seq4: D A A A
NAA = 14

14. MSA – Multiple Sequemce Alignment
An example
MSA – Multiple Sequemce Alignment
NAA = 14
NAD = 5
NAV = 5
NDA = 5
NDD = 8
NDV = 2
NVA = 5
NVD = 2
NVV = 2
PAA = 14/48
PAD = 5/48
PAV = 5/48
PDA = 5/48
PDD = 8/48
PDV = 2/48
PVA = 5/48
PVD = 2/48
PVV = 2/48
1234
seq1: VVAD
seq2: AAAD
seq3: DVAD
seq4: DAAA
QA = 8/16
QD = 5/16
QV = 3/16

15. Example continued PAA = 0.29 QAQA = 0.25 PAD = 0.10 QAQD = 0.16
PAV = 0.10
PDA = 0.10
PDD = 0.17
PDV = 0.04
PVA = 0.10
PVD = 0.04
PVV = 0.04
QAQA = 0.25
QAQD = 0.16
QAQV = 0.09
QDQA = 0.16
QDQD = 0.10
QDQV = 0.06
QVQA = 0.09
QVQD = 0.06
QVQV = 0.03
1: VVAD
2: AAAD
3: DVAD
4: DAAA
MSA
QA=0.50
QD=0.31
QV=0.19

16. So what does this mean? PAA = 0.29 PAD = 0.10 PAV = 0.10 PDA = 0.10
PDD = 0.17
PDV = 0.04
PVA = 0.10
PVD = 0.04
PVV = 0.04
QAQA = 0.25
QAQD = 0.16
QAQV = 0.09
QDQA = 0.16
QDQD = 0.10
QDQV = 0.06
QVQA = 0.09
QVQD = 0.06
QVQV = 0.03
SAA = 0.44
SAD =-1.17
SAV = 0.30
SDA =-1.17
SDD = 1.54
SDV =-0.98
SVA = 0.30
SVD =-0.98
SVV = 0.49
BLOSUM is a log-likelihood matrix:
Sij = 2log2(Pij/(QiQj))

17. The Scoring matrix A D V 0.44 -1.17 0.30 1.54 -0.98 0.49 1: VVAD
2: AAAD
3: DVAD
4: DAAA
MSA

18. And what does the BLOSUMXX mean?
High Blosum values mean high similarity between clusters
Conserved substitution allowed
Low Blosum values mean low similarity between clusters
Less conserved substitutions allowed

19. BLOSUM80 <Sii> = 9.4 <Sij> = -2.9
A R N D C Q E G H I L K M F P S T W Y V A R N D C Q E G H I L K M F P S T W Y V
<Sii> = 9.4
<Sij> = -2.9

20. BLOSUM30 Blosum30 <Sii> = 8.3 <Sij> = -1.16 Blosum80
A R N D C Q E G H I L K M F P S T W Y V A R N D C Q E G H I L K M F P S T W Y V
Blosum30
<Sii> = 8.3
<Sij> = -1.16
Blosum80
<Sii> = 9.4
<Sij> = -2.9