3

Brute force and motif finding

instructor: Ross A. Lippert

http://www-math.mit.edu/~lippert/18.417/

ctgttgaaagtatacaacatgtaagtctgttcatcttttcgtatcaatcg
tatcgcgctaaaaattatggtagttactaacgtagtggtatacataatgt
caactgccgcatataatggatttgcctagtgcttgaacggaggtgcaatg
atgtcaaaacgcctgaattaattggtaatactatagcggtgcggacccta
ttaagtattaggtgcgtaacctctcagggttgccgcccggttttatcctt
tgtgtaatagcctttttagagtcgaccgttcctcgtcacgcgtaaaattt
gtatgaatcctcgttggtttgtgggacgaccctttgtctatagtataaca
ccccggcaagttctaatcggctgtcagctactactatcctgggcgaacag
tgaaggcgtcgcgagtcttatgggtcaaatggccgaataaaacaatctta
tgagaggtctgtagacgacgattcgctgtcttatttgcccgccaagtaag

Lac operon in E.coli

Trp operon in E.coli

Trp operon attenuation

Eukaryotic regulation

But don't forget:

Chromatin structure
RNA transport/stability
Splicing

Some known Transcription factors

Factor	Sequence Motif	Comments
c-Myc and Max	CACGTG	c-Myc first identified as retroviral oncogene; Max specifically associates with c-Myc in cells
c-Fos and c-Jun	TGA[CG]T[CA]A	both first identified as retroviral oncogenes; associate in cells, also known as the factor AP-1
CREB	TGACG[CT][CA][GA]	binds to the cAMP response element; family of at least 10 factors resulting from different genes or alternative splicing; can form dimers with c-Jun
c-ErbA; also TR (thyroid hormone receptor)	GTGTCAAAGGTCA	first identified as retroviral oncogene; member of the steroid/thyroid hormone receptor superfamily; binds thyroid hormone
c-Ets	[GC][AC]GGA[AT]G[TC]	first identified as retroviral oncogene; predominates in B- and T-cells
GATA	[TA]GATA	family of erythroid cell-specific factors, GATA-1 to -6
c-Myb	[TC]AAC[GT]G	first identified as retroviral oncogene; hematopoietic cell-specific factor

Computational prediction of TFBS

upstream regions:

ctgttgaaagtatacaacatgtaagtctgttcatcttttcgtatcaatcg
tatcgcgctaaaaattatggtagttactaacgtagtggtatacataatgt
caactgccgcatataatggatttgcctagtgcttgaacggaggtgcaatg
atgtcaaaacgcctgaattaattggtaatactatagcggtgcggacccta
ttaagtattaggtgcgtaacctctcagggttgccgcccggttttatcctt
tgtgtaatagcctttttagagtcgaccgttcctcgtcacgcgtaaaattt
gtatgaatcctcgttggtttgtgggacgaccctttgtctatagtataaca
ccccggcaagttctaatcggctgtcagctactactatcctgggcgaacag
tgaaggcgtcgcgagtcttatgggtcaaatggccgaataaaacaatctta
tgagaggtctgtagacgacgattcgctgtcttatttgcccgccaagtaag

and a motif with logo

Motif finding

We have a collection of upstream regions and suspect a common binding site.
We don't know exactly where they are.
We have some idea on the number of bases in the binding site.
The string at those sites should all look `similar', whatever that means.

1/2*gtatca+1/2*gtatac
1*gtatac
1*atataa
1/4*ctatag+1/4*gtaata+1/4*gaatta+1/4*gtcaaa
1*gtatta
1/2*gtaaaa+1/2*gtgtaa
1*gtataa
1*ttctaa
1*gaataa
1/2*gtctta+1/2*gtagac

Collect the stats into a profile matrix

A	1	5/4	31/4	1	15/2	31/4
C	1/4	0	7/4	0	1/2	2
G	31/4	0	1/2	1/2	0	1/4
T	1	35/4	0	17/2	2	0

gtataa

One approach: finding similar substrings

---		gtatac			---
---				atataa	---
---	gaataa				---
---			gtctta		---

Find substrings with the `best' profile

Best profile = best CONSENSUS

Consensus

Select substrings

ctgttgaaagtatacaacatgtaagtctgttcatcttttcgtatcaatcg
tatcgcgctaaaaattatggtagttactaacgtagtggtatacataatgt
caactgccgcatataatggatttgcctagtgcttgaacggaggtgcaatg
atgtcaaaacgcctgaattaattggtaatactatagcggtgcggacccta
ttaagtattaggtgcgtaacctctcagggttgccgcccggttttatcctt
tgtgtaatagcctttttagagtcgaccgttcctcgtcacgcgtaaaattt
gtatgaatcctcgttggtttgtgggacgaccctttgtctatagtataaca
ccccggcaagttctaatcggctgtcagctactactatcctgggcgaacag
tgaaggcgtcgcgagtcttatgggtcaaatggccgaataaaacaatctta
tgagaggtctgtagacgacgattcgctgtcttatttgcccgccaagtaag

Profile/consensus

A	3	1	2	3	4	3
C	1	1	4	2	0	2
G	2	1	1	3	2	2
T	4	7	3	2	4	3

ttcaaa

Motif finding problem

Idea: find a profile which matches well to each string
Idealizations: we guess how long it is. Profile scoring is done according to consensus.
Input: sequences which contain a common binding site.
Output: the profile of the binding site.

Approach by brute force

Iterate over all N_{_1}*N_{_2}*N_{_3}*... starting positions S_{_1},S_{_2},S_{_3},...

What if we guess the profile string?

ttcaaa

ctgttgaaagtatacaacatgtaagtctgttcatcttttcgtatcaatcg
tatcgcgctaaaaattatggtagttactaacgtagtggtatacataatgt
caactgccgcatataatggatttgcctagtgcttgaacggaggtgcaatg
atgtcaaaacgcctgaattaattggtaatactatagcggtgcggacccta
ttaagtattaggtgcgtaacctctcagggttgccgcccggttttatcctt
tgtgtaatagcctttttagagtcgaccgttcctcgtcacgcgtaaaattt
gtatgaatcctcgttggtttgtgggacgaccctttgtctatagtataaca
ccccggcaagttctaatcggctgtcagctactactatcctgggcgaacag
tgaaggcgtcgcgagtcttatgggtcaaatggccgaataaaacaatctta
tgagaggtctgtagacgacgattcgctgtcttatttgcccgccaagtaag

Profile/consensus

A	2/7	29/28	61/42	1011/140	3163/420	788/105
C	47/60	12/35	381/70	1/5	107/210	17/21
G	223/70	16/15	23/14	107/210	31/30	11/30
T	2413/420	3173/420	307/210	869/420	389/420	277/210

ttcaaa

Note!

1+2+2+1+3+2+3+1+1+3 = 19

Median string problem

Idea: We want to find a consensus string which aligns well to each string.
Idealizations: same as the motif finding.
Input: sequences which contain a common binding site.
Output: the string minimizing the sum of alignment scores, i.e. consensus sequence of the optimal profile of the binding site.

Approach by brute force

Iterate over all [ACGT]^{^k} and align to all sequences.