Design Flow Enhancements for DNA Arrays

Design Flow Enhancements for DNA Arrays

Andrew B. Kahng1 Ion I. Mandoiu2 Sherief Reda1 Xu Xu1 Alex Zelikovsky3

(1) CSE Department, University of California at San Diego(2) CSE Department, University of Connecticut(3) CS Department, Georgia State University

Introduction to DNA microarrays and manufacturing challenges

Outline

DNA microarray design flowDNA microarray design flow enhancements:

Integration of Probe Placement and EmbeddingIntegration of Probe Selection and Physical Design

Conclusions and future research directions

Uses of DNA arrays

Introduction to DNA microarrays

Practical experiment using DNA arrays

DNA manufacturing process

Problems and challenges in DNA manufacturing process

Introduction to DNA Probe ArraysDNA Arrays (Gene Chips) used in wide range of genomic analyses

gene expression detectiondrug discoverymutation detection

Diverse fields from health care to environmental sciences

DNA Arrays are composed of probes where each probe is a sequence of 25 nucleotides

Images courtesy of Affymetrix.

Tagged RNA fragments flushed over array Laser activation of fluorescent tags

Optical scanning of hybridization intensities

DNA Array Hybridization Experiment

DNA Array Manufacturing Process

Very Large-Scale Immobilized Polymer Synthesis (VLSIPS)

Treat substrate with chemically protected linker molecules

Selectively expose array sites to light

Flush chip’s surface with solution of protected A, C, G, T

Repeat last two steps until desired probes are synthesized

Probe Synthesis

array probes

A 3×3 array

CG AC G

AC ACG AG

CG AG CN

ucle

otid

e D

epos

ition

Seq

uenc

e A

CG

A Mask 1

A

A

A

A

A

Probe Synthesis

array probes

A 3×3 array

CG AC G

AC ACG AG

CG AG CN

ucle

otid

e D

epos

ition

Seq

uenc

e A

CG

C Mask 2

CC

C C

C

CA

A

A

A

A

Probe Synthesis

array probes

A 3×3 array

CG AC G

AC ACG AG

CG AG CN

ucle

otid

e D

epos

ition

Seq

uenc

e A

CG

G Mask 3

CC

C C

C

CA

A

A

A

A

G

G GG

GG

A Nucleotide Deposition Sequence defines the order of nucleotide deposition

A Probe Embedding specifies the steps it uses in the nucleotide sequence to get synthesized

VLSIPS Manufacturing Challenges

Lamp

Mask

Array

Problem: Diffraction, internal reflection, scattering, internal illumination

Occurs at sites near to intentionally exposed sitesReduce interference

Increase yield

Reduce cost

Design objective: Minimize the border length

Unwanted Illumination and Border Cost

array probes

A 3×3 array

CG AC G

AC ACG AG

CG AG CN

ucle

otid

e D

epos

ition

Seq

uenc

e A

CG

A Mask 1

A

A

A

A

A

Border = 8

Border Reduction

Unwanted illumination Chip’s yield

Introduction to DNA arrays manufacturing challenges

Outline

DNA array design flowDNA array design flow enhancements:

Integration of Probe Placement and EmbeddingIntegration of Probe Selection and Physical Design

Conclusions

Previous Work

Border minimization was first introduced by Feldman and Pevzner. “Gray Code masks for sequencing by hybridization,” Genomics, 1994, pp. 233-235

Work by Hannenhalli et al. gave heuristics for the placement problem by using a TSP formulation.

Kahng et al. “Border length minimization in DNA Array Design,” WABI02, suggested constructive methods for placement and embedding

Kahng et al. “Engineering a Scalable Placement Heuristic for DNA Probe Arrays ,” RECOMB03, suggested scalable placement improvement and embedding techniques

Basic DNA Array Design Flow

Probe Selection

Design of Test Probes

Probe Placement

Probe Embedding

DNA Array

Logic Synthesis

BIST and DFT

Placement

Routing

VLSI Chip

Physical Design

Probe Placement

Probe Embedding

Probe Selection

Design of Test Probes

Logic Synthesis

BIST and DFT

Physical Design

Routing

Placement

Analogy

Design Flow Outline

Physical Design

Degrees of freedom (DOF) in probe embedding DOF exploitation for border conflict reduction

Probe Embedding

Probe Placement Similar probes should be placed close together

Constructive placement Placement improvement operators

Key DOF: Probe Embedding (Alignment)

A

A

A

C

C

C

G

G

GT

T

T

Deposition Sequence

CTG

Hypothetical Probe

Gro

up

C

G

T

Synchronous Embedding

C

T

G

As Soon As Possible (ASAP)

Embedding

C

G

T

Another Embedding

Embedding Determines Border Conflicts

A

A

A

C

C

C

T

T

TG

G

G

ACTG

AGT

GTG

A A

Synchronous EmbeddingA

G

T

A

G

G

T

A

Dep

ositi

on S

eque

nce

ProbesG

A

A

G

T

A

G

T

ASAP Embedding

G

Optimal Probe Embedding

T

A

AG

A

G

T

A

C

A

T

G

Before optimal re-embedding

A

T

A

AG

G

T

A

C

A

T

G

After optimal re-embedding

A

Using Dynamic Programming to optimally re-embed a probe

Problem: Optimally embedding a probe with respect to its neighbors

Kahng et al. “Border Length Minimization in DNA Array Design,” WABI02

AAAAAA

Placement Polishing Using Re-Embedding

Use optimal re-embedding algorithm to re-embed each probe with respect to its neighbors

Placement Objective: Minimize Border

Radix-sort the probes in lexicographical order

Probe 1

Probe 2

Probe 3

Probe 4

Probe 5

T A T T

A T A A

A A C A

G GC C

C G G G

1 2 3 25

T A T T

A T A A

A A C A

G GC C

C G G G

1 2 3 25

Problem: How to place the 1-D ordering of probes onto the 2-D chip?

Radix-sorting the probes order reduces discrepancies between adjacent probes

Placement By Threading

1 2 3 25

T A T T

A T A A

A A C A

G GC C

C G G G

Probe 1

Probe 2

Probe 3

Probe 4

Probe 5

Thread on the chip

1

2 3

4 5

Row-Epitaxial Placement Improvement

Array of size 4 × 4

For each site position (i, j):From within the next k rows, find the best probe to place in (i, j) Move the best probe to (i, j) and lock it in this position

Row placement = sort + thread + row epitaxial

Introduction to DNA arrays manufacturing challenges

Outline

DNA array design flowDNA array design flow enhancements:

Integration of Probe Placement and EmbeddingIntegration of Probe Selection and Physical Design

Conclusions

DNA array design flow enhancements

Integration of Probe Placement and Embedding

Integration of Probe Selection and Physical Design

Initial embeddings influence the placement results

Propose and implement two flows

Probe pools add additional degrees of freedom Integrate probe selection into physical design

Propose and implement two flows incorporating probe pools

Physical Design

Probe Selection

Design of Test Probes

Probe Placement

Probe Embedding

DNA Array

Integration of Probe Placement and Embedding

Probe Selection

Design of Test Probes

Probe Placement

Probe Embedding

DNA Array

Integrating placement and probe embedding gives a further reduction in border conflicts.

Probe Placement

Probe Embedding

Analogous to tighter integration between placement and routing in VLSI physical design

Integration of Probe Placement and Embedding

1. Synchronous initial embedding

3. Re-embedding using DP2. Row placement

Flow A

Row EpitaxialRe-embedding

ASAP initial embedding1. As Soon As Possible (ASAP) initial embedding

3. Re-embedding using DP2. Row placement

Flow B

010002000300040005000600070008000900010000

100 200 300 500 Chip size

Conflicts

6%

Placement + Embedding Runtimes

Row EpitaxialRe-embedding

0

2000

4000

6000

8000

10000

12000

100 200 300 500 Chip size

CPU (s)

1. Synchronous initial embedding

3. Re-embedding using DP2. Row placement

Flow A

Row EpitaxialRe-embedding

ASAP initial embedding1. As Soon As Possible (ASAP) initial embedding

3. Re-embedding using DP2. Row placement

Flow B

Second Enhancement: Probe Pools

Probe Selection

Design of Test Probes

Probe Placement

Probe Embedding

DNA Array

Physical Design Problem: Given a probe pool

for every target sequence, select a probe for every target sequence such that the total conflict after placement and alignment is minimum.

Probe 1 Probe 2 Probe 3 Probe 4

Gene Target Sequence

Probe Pool – Pool Size = 4

Integrating Probe Selection and Physical Design

1. Perform ASAP embedding of all probe candidates

3. Re-embedding

2. Run row placement selecting the probe from the pool that gives the minimum conflict

Flow A

ASAP initial embedding1. Perform ASAP embedding of all probe candidates

3. Run row placement using the selected candidates

2. From each probe pool select the probe that fits in the least number of steps using ASAP

Flow B

4. Re-embedding

Results (Conflicts) of Probe Pools

300000310000320000330000340000350000360000370000380000390000

1 2 4 8 16

Chip size = 100

Pool Size

Conflicts

2600000

2700000

2800000

2900000

3000000

3100000

3200000

3300000

1 2 4 8 16

Chip size = 300

Pool Size

Conflicts

8000000

8100000

8200000

8300000

8400000

8500000

8600000

8700000

1 2 4 8 16

Chip size = 500

Pool Size

Conflicts

1150000

1200000

1250000

1300000

1350000

1400000

1450000

1500000

1 2 4 8 16

Chip size = 200

Pool Size

Conflicts

Comparison of Probe Pools Flows

300310320330340350360370380390

1 2 4 8 16

Chip size = 100

Pool Size

Conflicts

2600

2700

2800

2900

3000

3100

3200

3300

1 2 4 8 16

Chip size = 300

Pool Size

Conflicts

68007000720074007600780080008200840086008800

1 2 4 8 16

Chip size = 500

Pool Size

Conflicts

1150

1200

1250

1300

1350

1400

1450

1500

1 2 4 8 16

Chip size = 200

Pool Size

Conflicts

Results (runtime) of Probe Pools

0

1

2

3

4

5

6

7

8

1 2 4 8 16

Chip size = 100

Pool Size

CPU (1000s)

0102030405060708090100

1 2 4 8 16

Chip size = 300

Pool Size

CPU (1000s)

020406080100120140160180

1 2 4 8 16

Chip size = 500

Pool Size

CPU (1000s)

051015202530354045

1 2 4 8 16

Chip size = 200

Pool Size

CPU (1000s)

Interpretation and Summary of Experimental Data

Initial ASAP embeddings produce a decent reduction in border conflicts.

Probe pools offer an extra degree of freedom exploited to further reduce border conflicts

Integration of placement and embedding yield up to 6% improvement

Probe pools add an extra 12-13% improvement

Total improvement up to 18% compared to results published in the literature

Open Research Directions

Probe selection should incorporate ability to uniquely detect target sequences present in sample. This should be done with no ambiguity. Methods similar to Boolean covering and test diagnosis can be used.

P1 P2 P3 P4 P5

T1

T2

T3

T4

Each target sequence should have a unique signature

Probes

Target Sequences

Open Research Directions

Stronger placement operators leading to further reduction in the border cost.

Insertion of probe test can benefit from test and diagnosis topics for VLSI circuits.

Future work also covers next generation chips 10k × 10k

Conclusions

We presented a DNA design flow benefiting from experiences of the VLSI design flow

We introduced feedback loops and integrated a number of steps for further reduction in the border cost and hence unwanted illumination

We examined the effects of probe selection on both placement and embedding

We examined the embedding options and placement on the total border cost

Thanks for your attention