Real-Time Primer Design for
DNA Chips
Annie Hui
CMSC 838 Presentation
Use of primers in PCR and Microarrays

PCR (polymerase chain reaction:

to amplify a particular DNA fragment

Use: to test for the presence of nucleotide sequences

Test of PCR products:
Ladder: a mixture of fragments of known length
Lane 1 : PCR fragment is ~1850 bases long.
Lane 2 and 4 : the fragments are ~ 800 bases long.
Lane 3 : no product is formed, so the PCR failed.
Lane 5 : multiple bands are formed because one of
the primers fits on different places.
CMSC 838T – Presentation
Use of primers in PCR and Microarrays

DNA chips (Microarrays):



to analyse a large number of genes in parallel.
fluorescence
Primers:

20 to 100 bases long

Synthetically manufactured
Bound to primer
Automated design of primer

A computational approach

Objective: To find primers that bind
well without self-hybridizing

Critique: how accurate?
Fixed on chip
CMSC 838T – Presentation
Motivation:
This group uses the
automated NucliSens
extraction system
(bioMerieux) to
develop their primers
here.
CMSC 838T – Presentation
Technique: The computational model
1.
Select primers from target sequence

two primers P (forward) and Q (reverse) for PCR, one primer
for DNA chip (microarray)

Using window size W, number of possible primers with length
n
between m and n within 1 window is:
S  l m (W  l )  1
CMSC 838T – Presentation
Technique: The computational model
2.
For each primer pair, or single primer,
Quantify 4 hybridization conditions:
a.
Primer length
b.
Melting temperature
c.
GC content
d.
Secondary structure
i.
ii.
iii.
iv.
We are starting here
Self annealing
Self end annealing
Pair annealing
Pair end annealing
CMSC 838T – Presentation
Technique: quantifying hybridization conditions
a.
Primer length len(P)

b.
Affect melting temperature and hybridization
Melting temperature Tm(P)

Temperature at which the bonds between primer
and gene sequence break
n 1
c.
H  p    H  pi , pi 1 
CG content CG(P)

T
 p 
H ni1p1 
m ,1
S p  S  p ,p
G-C pairs are more stable than
A-Tpairs
S  p   
R  ln  4 
(because of more H-bonds)
p  primer
i 1
i
R  1.987cal / C  mol
  50 109
# G in P  # C in P
GC  p  
100
T0  237.15 C
p
What is this measure good for?
t  21.6 C
H  p   enthalpy
S  p   entropy
CMSC 838T – Presentation
 T0  t
i 1
Technique: quantifying hybridization conditions
d.
Secondary structure

Study how likely a primer entangles with itself or with another
primer

P = {p1, p2, …, pn}, Q = {q1, q2, …, qm},

Scoring function:

S(pi, qj)
Example:
= 2
= 4
if {pi, qj} = {A, T}
if {pi, qj} = {C, G}
=
otherwise
0
Position i of primer P
P: ...AGCTTTAGCCATAG
Q:
TCTTAGGATCGC...
score S(pi, q1) = 2+4+2+2+4 = 14
CMSC 838T – Presentation
Technique: quantifying hybridization conditions

Four measures of secondary structure:
i.
Self annealing,
•
SA(P, P’)
P’ = reverse of P
SA( p, p' ) 
P
m
max  s( p , p
k 1 m ,..., m 1 i 1
ii.
ik
')
P’ P’ P’P’ P’ P’ P’
Self end annealing, SEA(P, P’)
•
•
•
iii.
i
Like Self annealing
P’ P’ P’ P’
k>=0
Only count longest continuous overlaps
Pair annealing,
PA(P, Q)
P and Q are the forward and reverse primers
Pair end annealing, PEA(P, Q)
•
iv.
P
•
similar to self end annealing
CMSC 838T – Presentation
Technique: How to apply the model

For PCR:
SCPCR( p, q) 
[ len ( p) GC( p) Tm ( p) SA( p) SEA( p)
len (q) GC(q) Tm (q) SA(q) SEA(q) PA( p, q) PEA( p, q) ]

P is forward primer, Q is reverse primer

Ideally, no annealing, length, GC and temp of P equals Q
SCPCRideal  p   len p
0 0 0 0 
w   0.5 1 1 0.1 0.2 0.5 1 1 0.1 0.2 0.1 0.2

GC p Tm, p
The optimization is:
0 0 len p
GC p Tm , p
min lPCR  p  
p
where
lPCR  p    SCPCR ( p, q )  SCPCRideal  p    wT

For DNA chips (Microarrays):

Q doesn’t exist. No pair annealing to study. Only 5 terms left.
CMSC 838T – Presentation
Technique: parallelize SCPCR(p,q) calculation
Compute PA and
PEA in parallel
Calculate Len, GC,
Temp, SA and SEA
in parallel
CMSC 838T – Presentation
Technique: details


Melting temperature and CG content:

Simple adder+divider

Use pipelining

1st one: O(m)

Subsequent cost: O(1)
Whole window: AGCGATATA
i-th P primer:
GCGATA
(i+I)-th P primer: CGATAT
• CG(Pi+1) = CG(Pi) - 1
• H(Pi+1) = H(Pi) - H(GC) + H(AT),
• similar for  S
Annealing matrix
c
b cd
a bd ce
ad be cf
d ae bf
e af
f
CMSC 838T – Presentation
Complexity

Complexity for sequential algorithm:

For PCR:
p
Number of choices of P (window size=Wp): S  l  m p (W p  l )  1
n

Number of choices of Q (window size=Wq): T  l  m (Wq  l )  1
Each distance SCPCR(P,Q): Ol p2  lq2  l plq 
Total: OS  T  Wp2  Wq2  WpWq 
nq

q



Complexity for parallel algorithm:

For PCR:


Distance measure SCPCR(P, Q) = O(1)
Total: O(S*T)
O(S*S*T*T) is a typo in the paper
Similar but simpler for Microarray
CMSC 838T – Presentation
Evaluation


Experimental environment

512 primer pairs, |Wp| = |Wq| = 16
1.
500MHz Celeron system with integrated hardware accelerator
2.
Software implementation
Evaluation results

1920 secs for software implementation

3.41 secs for using hardware accelerator
CMSC 838T – Presentation
Related Work

Previous approach

DOPRIMER




Same computational model
Differ in the way of doing dynamic programming
Sequential in nature
Other Primer selection softwares

Eg: Primer Premier 5, Primer3, PrimerGen, PrimerDesign

Similarities:

Criteria: Length, Temp range, GC range, GC Clamp, 3’ end stability,
uniqueness of 3’ end base, Dimer/hairpins, Degeneracy, Salt
concentration, Annealing Oligo Concentration, etc
Differences:




Not a weighed linear sum of all criteria
Need much expert’s supervision,
the numerical criteria are used as a guide only
CMSC 838T – Presentation
More Related Works

Case study

Burpo did a critical review of PCR primer design algorithms


Subject: saccharomyces cerevisiae deletion strains
Conclusion:


no suitable program for the task of post-design PCR analysis
Especially in the aspect of accurately predicting non-specific
hybridization events that impair PCR amplification.
CMSC 838T – Presentation
Observations

My observations:

Minus side:

Is the computational model too simplistic?
 Specifically, is a weighed linear sum justified?
Plus side:

The design of the parallel architecture is neat.
 Since primers are about the length of 18-22 bases, current
technology certainly can handle it.
When would you need fast primer selection?




Primer walking to connect contigs together quickly
To scan through a large number of sequences for possible
primers
CMSC 838T – Presentation