9002 scr

Genotyping Polymorphic Bands of Microsatellite DNA withAutomatic Object Segmentation and Tfanslated Absolute Size

Chun-Fan Chanprr'-. Chang-Chih Shent'. Chong-Long Huang3, Hs_ueh-Ting Chuar, Hsiao-yun Kuor,Mei-Ling Wang'. Wen-Hua Lin', yueh-Tsu King6. Chih-Fe^n^g Chen7. yen-pai t_ee7. Cheng_yan Kao3.8,

Chaur-Chin Chen'"r

Ab6lract

Automatic image data analysis (Aida) on the Senotyping potymorphic barids of microsatellite DNA re4uires routinelvaurormtic image segmentalion and especially translated absolule data. Compulerized aulomatic analysis is crucial lnreproducibility and artifact-proof despite of the available connercial software which is highly inreracrive and hence gravelytirne-consuming Albeit, the genotyping data of microsatellite DNA fragment that is represented as relative nobili[ sizeswith floating points based on selected DNA size standards may lead to great loss in dattintegration. The implenented Aidasystem with general-Aida-modules (G. rf) of object segmentation and specific-A,da-modules (S,4M) of data translationintends to provide efficient solutions iD its semi-automatic format. In an experiment of twenty images, tIrc Aida systemdemonstrated 85-95E accuracy in automatic band segmentation. Moreover, the Aida system alo;g wi0r in-house DNa sizestatrdafis of Ma*Qoff Marker-Quarter-oif) successfully translated absolute size data of DNA fraanents with zero standarddeviatlon while analying lnplicale genotyping images of hree running distances. In contrasl io Arda system's superiorperformance, the results of conme.cial software based on fifteen DNA size standards in triplicate genotyping imagesdelivered the DNA fragnent size data up to 2.19 base pair of standard deviation along with the diicrepancies oi dati noatingpoints and missing bands.

Keywords: Automatic Irnage Dada Analysis, object segmenralion, Absolute size, Band, Lane, size standard.

I Breed-Use-special Laboratory, CeDter ofAgriculture Hierarchical Utilization, C,raduate lnstitute of Biotechnology, CollegeofAgriculturc, Chinese Culture University, Taiwan; *, first author'Department ofComputer Science, National Tsing Hua Univenity. Taiwan; *, fi$t author ofequal contributions; g.corresponding aulhor.'Departrnent ofComputer Sctence and Information Engineering, National Taiwan Univenity, Taiwan." Department ofComputer and Information Engineering, Asia University. Taiwan.'Animal Health Research Institute, Council ofAgriculture, Executive yen, Taiwan." Animal Technology Institute Taiwan, Taiwan.' Departrnenr ofAntmal Science. Narional Chung Hsing Unrversiry. Taiwan.'Algiogenesis Research Cenler, Nattonal Tawan Universiry. Taiwan.

1. Introduction

Genotyped polymorphic baDds on bio-images ofelectrophoretic analysis are represented as relativemobility sizes of DNA fragments based on selected DNAsize standards. Polyrnorphic mobility patrems ol interestedband objects (BO) are often analyzed among individualsubjects with respective DNA samples to comparegenoilpic relations. Importantly, relative mobility sizedata with floating points has imposed serious obstacles togenotyplng among data integration, procedurestandardization, and system automation with availablecomnercial software packages [1-4].

The bio-images are with sirnilar inage objectsproduced by biolectmiques with practical purposes ofgenotyping, microarray and two-dimensional gel. Indeed,indispensable image data analysis software bridges fterevolving bioinformatic daia analysis and bio,experimentaltasks. Wilh dynamically ad.iustable intensity range formarking BO from local background, boundary drawing ofBO is cmcial for accurate genotyping jmage analysist5-el.

The basic modules of object segmentation and datatranslation in automatic image data analysis (Aida)algorithm demands complete implementation to meet thepractical specifications of general object segmentation andspecific task goals. Segmenring BO of microsatellitegenotyping images essentially deals with contrast quality

of local background and objects. With dynamic thresholdbased on contrast quality, object segmentation sansitivelylocates object boundaries better than the cases based onsimple average filter of noise elimination especially withthe cases of spread-out and blur objecrs [10, 11].

In this paper, we implement general ,4ida modules(GAIO of object segmentation to reach the 85-957oaccuracy in standard proc€dure with 20 inuge sampleswith minimal user efforts of sefting 2 parameters. Inadditior, we implement specillc Aida nodules (SArlt) oflane finding and malker localization to translate theabsolute size data in nucleotide lengths. Specifically forgenot)?ing tasks, GAM computes single band wilh obje€tsegmentation and SAM respectively computes goupedbands with lane finding ard marker localization ofabsolute size standards to translate absolute nucleotidenumber of microsatellite genotyping fragments n2,l3l.

2. Materials and Methods

2.1: Microsatellit€ Genotyping Analysis

L Microsdtellite DNA AmplifrcationThe genomic DNA samples of total 6 chickens areanalyzed with ADLo102 IRDye700-prirner (Tm 54.3"C) arregular PCR program, respectively. The PCR reaction iscarried out under 36 l(denature), (anneal), (react)] cycles at[( 'C. second)] conditions of [(95,30), (52,15), (72,30, andinilialhot stan. The ADL0l02 lRDye70o-primer is from

- 1324 -

- szel -

atnlosqe Supuodsa!oc Jo laqunu eqt uSrsse ol (SAldaruanbas go1ryow qu^l poualp aJu oNlo8 a^rlJadse! 's8uuts-iJ8 uee,,\\la8 :ra8alur (l+I) ot lEnba aJuetsrptlun llgy Ja^o acwlsrp d?3 Jo onrJ eqt ,4q peuruualapsl Squts-&g jo de8 1g roJ ra8alur-l aql .?Jrr?tsrplrun rep8a: uerl arwlsrp papualxa q1r,n de8 ggluarrfpr 1epouasur (OrV'trafqo euou) N *l s? pw (OB) I sE p?ar ereeuul pnjopusls dzts Jo og possorr,?Jt{ aqJ ,oIl?J 77n)7Jo 60 ueql lol€er8 rlft\ pu? qt8ual ZJ? lsa8uol qtr^\auel p[jgpuets ?zts aq sr (7)W) sauj Jatua) Dr[rDW atqJ-

Ioltolsuo,u puo uoltmio.al n\Jon .€'euBI a^!.adsal Jo 7J? ,{q passolJ lou aJe

qcnl{ spueq asrou IIErus lJo rallrj ot Japro w gg se slulod797 saqddr (u?ls,{s ,pry aql .sauel uaa^\taq Eupcaualur^w tnoqlr,r\ (Z lo Z.l.l-=u) yJJT Buuoqq8rau ̂\auolur 1;1qs ,{nu uoqcalp srx?-,{ ur u,q41gy ?J7 qtr^r seuultuecrfpr, o,rnl Jo spuaJl Jelrruts eqt .stJefqo auel padnor8Ala qJadsal Jo 7Q7 pu e 7J? papo^al qlta .elDrrrploo.^\or Jo u qlr^! (w ,u)X.Bqrcqt$pN

Jo sp (u.u)laxrdol lsasolJ st qJlqa aleurpJom-x gg ,tq las aJe sauepunoqawl aqt'sqtpft\ aull snous^ aql Squapuo:q Jo BulJnpaJplo^? oJ elEutplooJ-f, j)j a\r ulo:g ,{1pra1elrq ocwlslp,UgV Jlv\ se pelndruoc erv sauepunoq tq8u-pu?-Uallllpal eql sDeJaq^\ awl aldrurs Jo sauupunoq tq8uqw ual aqt saur.Jap 7c? luJal?lrq eql leurl

]o sraluac pueqduplull 7J7 rlr.n. spurq passoJr 7J7 s3v9l loqlo wo{apls? elqlssod sr spueq,{u?ru ss ssoJc ol Japro ur adols 737u spual sleu?l ot Surpro.r st 737 Furpuelx3 .uotduJsep;w^lolloJ orlt uI patsrt sr uqtuo8p eqI .797 Bulddel:a,ropatf,asretul ou lnun ?17 pan?d Jo Swsserold a^rsnral JaUertd 75-7 Stnddrlre,ro Jo 75? pasq er11 sr ?J7 aql .aJnr;dul "Juetqp x Jo ploqsaJql JrurBu,{p luarcr.JJa u? slptsutHBV*'I a\I lHB jqStaHpuDqaSora^V *-I ro A^gVueqt ssal ,(la^qradsal sr dJ Jo eJualaIlrp atpurpJooc-F^aql Jo acu€tsrp-,{A alqpu?^ eql Jr ?JqeBF J?autl Jo se u t?lpsr dJ a^rlmasuoJ aaJqt ,{u! Jo (TS, saur.Iannbsjso4 V

uoucnJtsuoJ puo 8r!2poJJ arvl ,z'eldwes 8urd,{louaB

a,rqcadsar ;o st efqo podnoJ3 aqt JoJ s€ pueq IEqrJouqll,{ patalduro. sr uonetuaruSes lcefqg.suoqrsod auelur sa:{els(u plo € ol Og IBruJou ur salrr.pJooc_,{A qloq Jouorluwtuns slaxrd plol Jo ue?u aql sr dJ Jo elrurproor,{7xarll \ABV) l|iptl puoBa?Dra^V v!-r! .Ie88rq sr qtphr pueqJI palalap sr dJ Jo uortrsod lalual aql .atEurpJooJ-^ pu?apurpJooc-x ruo4 ,{lalucodsar slaxld euallxo Jo acuoJaJJrpalnlosqP aql aJP 1q31aq pue qlpB\ pupq aqJ .spuyqpulou Jo se Hh,sv lo ploJ-z.l pup pllttrl ?uo uaa,rllaqsr raqunu slaxrd papuno.uns qtl^\ ezrs pu?q aql alrq^1og lo ratua. eql sr (dJ ) pxldtatua) aql.^llvralx|lq (7Q7)sautTuouo rrwa1autl ot 1 pw (I)1 aurlJatuaJaualauo qtr^\ Og 8uld,{tou?8 Jo ewl aldtllEs uo s]cerl /,{yS aql

lotlua) aurl .I

tr\lys ,{q uoqBlsup{ lJafqo padnoJ9 :[.2

lnvg srlnpou t plv x{pads lo yeqr aqd .Z arn8li

.uofal purq Surpuods?rJoc oqlur palcoiap slexrd Jo 1e,re1 ,{rr3 aSera,te aqj qllt\ patprJosseproJluao aql ,{q petuasarda: aq wc lcefqo pwq qf,ua .sJellrJ

uerpalll Jo u?atu qr^ Sursn ,{q pol?u(urla aq uBJ asrouaqt ,{lpurd spwq luacefpr o.n1 pu? sauel Buuoqq8rau o^ilqtoq uaa^\]aq saJuelsrp lrcrs,{qd Sullndruoc ,{q pa,rlos aqwr spu?q pur sauel SurddelJelo .JaqunC auel lrnpt,ttputqcza uo s1a,ra1,{rr8jo urtJSolsrq aql uo paspq uorpluau8esaSrur fldde a^\ 'auel qrea JoJ uaql .uotnqutsrp (,{)HJo s,{allu^ aqt o1 Surptocoe seurl IEJrua^ aq1 aluzdas pw

'J>r=0 ' (tx)17+=(,()H ,(q uortrarp lquozuoqFX

aql otuo slaxrd aqt lJ?[oJd lsrJ a^r .l ern8rg ur a8euttndu aql qlr..l sanbluqcal uol:elueuias aiutut ,{e,r-o,ruuo pasEq a8nur aql w (suorSal Jalq€uq .Og) slJatqoprrBq aqt Jo llt atuJol ot sr po8 eql .suunloJ

J pu! s^\otU Wra a3?ur 1aB lndur w aq,J>,{=0,U>x=0.(I.r)I lsl

(O&) qza[qo puDg atDxul

,^lyC ,{q uoqquaur8as Da[qO aFurs :Z,Z'uoltrslnbrb

O& (q) puv,etolat p3 pa8uo @) INVC)salnpou oplv ptettaS WA (O st,�a{q9'I etnE!!

(q)

'suorpes nuhrolloJ a[D rrr alPslrytap uonquaueldul aql qJtq^t Jo aJr^ruos ,pry asnorl-urqll/r\ poz,{leuE s1 epp aSeur CCIJ llq-91 aqt la^oarol,\tr 'azrs {lllrqoru a^lelal Jo ettrp slqod Btntto[ Eurla8roJ goo: ?l{ pur vovs r{loq Jo uoqpuuoj(Ir pJspu?lsazrs aql Sulas ,(q (VSn..rul $rl,i{ peqddv) aB?IcBdaJs^ruos scueunNotg prerauuoJ qlm paz,4pue sr eppafeur 8qd,{1oua8 eql .lspJ

t Jo paads aq1 1e las JarruBcsrasel aq1 ,{q euoz-pue IaB Sulqr?a uodn ellJ CCII llq-91se paqdec sr sptdrs VN( paruud-6g1a,{q11 3o a8etur1aE aql unr3ord 1p.n 0t puo.epraduregz tlo^ O0Zl qtft\uorlrpuor Jogt lr srsa:oqdo:pa1a to; palfslsur sr Earll qlft\raJnq AgI x1 ur pB apnuel,{rcu pozuau.{lod %S.g Jo ta8urtu-ogzxz 0 aql ,{la^rpedsal ,pawol

luualetll .ln 9 0 qlhr(VSfl "oul

UOJ-I.I) reruanbas-otnp OOZ' ualsr(S pqolCuo petwsd€s an Soj4.toy1 pus ygy5 Jo saldruzs :a1:erupur saldues Surddloua8 Srnpnlcur spualuru paqruapaqt 'seuEl lat Eur l?^latur qaa^\Fq-(II satrel aldr[Bs OI qu^{

aroqlos onv puo s.uaumNoq :qs{louv azouq t '(VSn'cul UOJJT lraraq paslqund

sr dq 0s€ pu8 'szE 'mE .997, ,jez 'toz 'Nz ,gLt ,qtl'0Zl '90t '001 't6 '91. '09 Sulpnleur prepuEts n$ V)VSIll aluuoc'uoqrpp8 uI .(VSn ..cq eIu8rs) JeJJnq al?F Eunluoururrl,{qtaul qlvi\ (ued?I 'cul rr{c€trH) uoqecgundJ-IdH 000t-c roJ puu (vso 'oJ gulNgltdg)t[) II }tr'IAJXg ruroql mbas qtk\ (dururaD , oJ O,/!\K)Jalc,{rouuaql UJd 96 snupd loJ srseqlufs uolsnc uollsr:atuud-gg1ax6;1 /I aql uEts roq pnrur qrr,.ir [tg6'91r '(St'09) 109'96)l Jo suoqrpuoc [(puocas 'Jo)] ]t sap,{.[ocpa]) '(lEauue) '(alnlsuap)l

99 rol :aruud-gg1a,t6g1/J qtr^t paruenbes s1 apldual SAW aql .KOW)aJuentlas JJobrlJ8l^l u^roul qll^! alBlduot S6cuanbaspallooe aql uo peslq spJyprrEls azls atnlosqe peqclJuaarn aJ? ple Surruanbas dI(C.J.V)pp paxru Jo (dq0OS't ol 0E) sluauS?)!VNqgoAyolAt parlund-J-14g aq1

tIoAVDn * VDVS :sptopuqs a\S an.lard Z '(AUP(ul.D'OJ

Dldl^i) Jap(roruloql dJd 96 sftuud loj srsaqlu.{s urolsnc

J

sizp. Ttte MQS is translated as B at ACG atrd as .lV at Ttowards ,lv-string. Tl]€ BONO of identical absolute sizenumber anong marker latres are grouped by respective lanepositions and are linked-up as horizontally parallelAbsolutesi?tLinc (ASL). The absolute size of sample laneBO is assigred based on the ovedapping AJr.

2.4: Datrbase of Aida Systen

l. Gewtlping Resul/J' (GR) for Ai.h DatobqseMa*c t l ,qne2031Samplz la te456789Lan 0 216 217 219 220 222 223 224 225 227 228lanc 0 230 231 2i2 2i3 234 ... ...PIXEL2 1 10n 17 8189 I I 217PUEL2 I IoM 17 6930 0 I 217PD(EL2 I 1086 18 7968 0 t 217PIXEL2 2 1105 16 8957 I I 219PrxEL 2 2 tIA 16 8538 0 I 219The GR text file saves the Aida aDalysis results of BOiDcluding band number, m8lker location, and band intensityfor loading into Aida database.(1) MARKER I-ANE (M.), SAMPLE LANE (Sl), Lane

and PIXLE are the record types of clR shown at the firstwod of data line with TAB separator.

(2) MARKER LANE records marker lane indices.(3) SAMPLE LANE records sample lane indices.(4) LANE records lane inder at second word aloog with

maximally 10 size marker locatioos. The ML or SLrequircs nxmy Lane data lines to record all the markerlocation of BO.

(5) PIXEL records lane index, band index, x-coorditrate.y-coordinate, intensity, band center, marker lare, andmarker location of 80.

2, Genowing A,lltsiiBy ttre cR example, 4 Ml and 6 Sl arc shown with mar*erlocations at 0-th lane including 216,2l'l,219,22O,222,utd so forth. PXEL [1087,1fl is the center point of lst BOat 2nd lane of matter lane with inlensity of 8189 andabsolute size of 217. For the genot)?ing analysis, the G,Rdata is loaded into Ai.da database including sample laneswith band absolute size and band intensity for the bandhtensity c{rtoff ratio in each latre along with detailedexperiDetrtal bformation.

3 Results and Discussion

3.1: Single Dand Object Segnentation

Our Aida syslem applies local backgrcud threstroldand band size respectively to filter nois€s of low-frequencyand high-frequency automatically. The seguEnled object ofsingle genotlping band is sirowtr by t}re band boundaries(88) of eremplilied band object (AO) in Figure 3. The lare80 is efficiently located with flexible 88 in overlappingstretches of BR as shown in Figure 4. Desoiplive detail$ ofthe algolithm are at section 2.2 of general Aida mdule(G,AM) on single object segEentation.

Our Aida software nay wort more accurately withrealistic genotyping bio-images of slab gel electrophoresiswittrout lorowing band regions at first of whidl is preliselythe idealized prcvision in cornrnercial image data analysissoftware of genot)?ing gel images. Specifically, our Aidasystem intends to r€solve the issues of excessive natrualinteractiotr atd accurale band segmentation by autonaticprocessing with only two parameters of local backgpundthrcshold and band size entered by user. Additionally, the

peak, valley, band size, aDd local mean of pixels intensityare also included for automatic band object segmentation inAida system in order to rcctify the fact that these criteriaare rtot universally efficient due to the required provision ofideally uniform band object and lane.

Comrnercial softvare for gel image often requires tomanually set lane positions first before band segnentation.Moreover, some software may overlook image proc€ssingissue with only vertical waveform peaks to segmeDt bandobjects. Especially, some software may mistakenly segmeltband objects by presumed regular shape of rectangle in tlrcgenotyping gel inage.

Figun 3. (a) DO l,nq.O)BO tcgrn.ntdtioD nith BB is tet dt r.d pixels. Th. BO and BBtbn^ Jor bard frj.ct and bdrd boanfuf.

BO it sel b! tctlav.greenkh p:rtcb, The BR is bdnd ngion.

3,2: Gmuped Lane Objects Translation

With specific Aida modules (MM) for polymorphicgeDotyping images data analysis, band objects in r$pectivesample lanes arc showl iD Figure 5. The batrd boundary(BB), lare demrcation lines (lDL), lane cents tre (rcL\,and absolute size lines (ASa) are autonatically idertifiedby the descriptive detr|ils at sectioD 2.3 of grouped objecttranslation. The microsatellite fragm€ots of individualgenomic DNA sample arc translated into absolute sizenumber as of grouped lane objerts based on ASL arnongMorkQof leuJes which arc often rerognized with longestimage coverage.

Ou! Arida algorithm guarantees that all lsl acrossrespective lane cetrter with bolh .4Jwlimited la[e CP andL*ABfilimited stuaight laDe. Greater L value may be equalto the trumber of lane objects crossed by the straighler rCZin lalle tracking. The lare tracking methods may resolve theskew issues of object lanes aDd genotyping image. Inaddition, the S/M noise filter herein by both band size andlane position may install appropriale Aida procedure forgenotyping tasks and rescue mis{eletion on meanilgfulweaker BO in whjch occur often at initial analysis stagewithout valid certification.

33: M icrosate[ite Genolyping Analysis

The Aida system and MarkQoff size standards hassuccessfully translated lhe absolute size data of genotypingDNA fragments vdlh standard deviation at zero base pairby anallzing the triplicate genotyping images of identicalruming distance in Li-Cor Global System 4200 (FiSure 6and Table 1). With same geDotyping images, commercial

o)(a)

- 1326 -

- LZet -

'saEr{[r rnuolducs(rul prrE crtllorc8qloq q poldd? spoqFru eqt auguoc atyul rlu@lord Josls,(t uB (llu8uodur eJoj^l slalo.d aJn]€al pelslar-ou?ueJssB ftpcFsnqs ptzIIBuB a,3 laDuoo pquou uo p6sBq onf,lu6ouo-dn qcJq[ dq slrmo(uB uralad snout^ uodn puErsSqqnts Jo eSEqrr spds pJtoqssaqJ ouFuroaE sl?e,\al latlBuorsuaurpo,tll JuroeloJd .aua8 eJnpal pelElaJ-otr8uacssB ,(llsJFsrlsls pazi(puu arr lq4uoc FurJou tro pesEq orlBru,rop-.ro-dn qcJq,t ,(q qa,ra1 VNUur snoFt^ uodn putrsuoF84puqdq Jo a8rut slop pJroqsserp [BuoEoquo sFa^aJduleoJcnu cFrotdF"suu{ .acuulsrp qpua8 sr fllscusnqspolJllrBnb aJB q€lqns 3o uonela: crddlouat qclqa Iq sezrssl[aur8uJ snog?A uodn srualed ,(11lrqour yo a8uul spwqqqd:orudlod qra,ral 1a3 Erqd,{t ctruoue8 ,praua8 uI

Jatpq eql pur sluorutar3 SrldflouaE.;o rlup ezrs atnlosqErqsuapBJ?qc aqt seuj?r asBqBtrp rply aql .1Ia pue vgvloJ anlra e.Iuru ,{lqtssod 1o aq i(uul r{cFlr! 14 uortBr8alurrltp pug uorlruotna uals,{s uo e.utuuo}ed ratFq i(lalllJo uatsi(s ,pry aql tuaualdul a^| .Jadtrd sm uI

uorsnlJuoc n.(TIb) I,ol lrEJl a^qqnE?nb

pu? (VCV) srs,{Fue Eud{loua3 4lalB Jo sB s:1su1 paqoadsJo srs,{lBrrs BlBp luenbesqns s-aurF[?a4s pue sauolsoqrlprrB 'sosBqqBp 'quauuadre luaJaIp Jo uollElEalur ElEppcp*rd sayquua Og petdaccB Jo nqurnu qlSuq alnlosqepolBlsrrD4 aql Sqssaoo.rd trtzlplrpusts puB crEurotnp uracwldarJB Og JoJ oIrJ JJolIo Eulzlupdo spt.rol {usualulspafqo auEI arnoe loJ /{lrsuel(Il Og f8npl^rpul dn srllnsJequrg ualsi(s tprv eqt 'as8qrl8p EplV SulpDol rod

'sa8ruu 1aE pepolslp yo IraJJe alrurs pue

sawl poloorJ 3ur^tosal w acuuu:opad aremgos rouadnssu sal?Jlsuqnep .{puaorga ura1s,(s egy fq 6991-96 1o,(rEJncrr pe^alqJe eql uowlsueJt t]lp e4s alqBldeor€eq lpsal uaryo deur rpq^r acue6p ,(111qoru papuo[a pu?tsJJa alrus sa^losal fDualcrJJa uels,(s BpJV eqt .saBBr.[ruI saFulpJooo-,{A uEql Jaqlsr acrrBlslp pu?q Auuns?alll dg 'souruacs ifio1e.roqt1-Jalur pca .a88urFJalur .aldturs-.tatw

Jo uorlEJ8etul Bl8p elqunlB^ aql s.JrrEque q]Eual luau8E{Og Jo rlep az rs alnlosqs Jo ualsds BplV aql Tfs€ O 'sasrc ere,ngos rulnSar (II uollElslr?4 rlrp e4s OgsnoauorE esoBc ,{ErrI plag rtJ|Jala IaA rBInBaJr Jo pelJaelFus uoltBr^ap pJepuEts JatEaE olul ulup ezs ,{1l1qoure^$?lal aql El4lcaJ$ pue alqeuer' aq i(yru sp.rBprr?ls ozrs Jo,{lqlqou aql qrru,$ q1r.r acurlsrp Suruutu elqrugl Jo rolclJasJa^p8 [ruorUppB [slsur f?ut lat Sur'uenbos l€uorl(Er^uoouo SuldfloueE Jo ru?l8olplJ ftaln mareq perrdruoc ssulcaru,rgos .lrp8ar aql ,(q dowda.nsp aql sapJolletur rlcg1tsJI?d esBq EZI-ZII Jo a8uar [Fru$ aql urlllur a.n sluarutu:3SurdfloueS Jo $aqunu ?zls palD[sqBJl aql flqeloN

'tta.g.a&u tsexg e\s 4nptqo ptu ,eul ratneJ euDI 'tarn ,.ogouort p .rrot .ttqrrnoq Foq ut 1SV ptto 1n 1@ 'AA eqJ 'tarn ,taed u! qe4d ttlt ,o t t t! ?SV .ll .ta,/:i edrntAon.t ptt 'fitq ?er 4 sp4d .r!t tt tas e.o.IJtI prn 1@ ,Ag

reuud-46efuU ZqItrIOV tt ?a p.{ttd@ sutryyzt ro"g#otttdwt Ettd&mat qnaoso,tru 4s aro 9-I sautl ,cpmp@tteztt VOVS pto ffipn o r.Mt S pw n .ql ,n pr@ ,S ,9'S , '€ 'Z ,1,

S ,n clo .3ot4r t dtet qtue to s.rtut .ttt ..to l[os

tttlaunNq& (q) .8ottt nol @) .Wpntt .zfs VNO gqyonarno+,,|p@ ut4tfs qV f4 tat',orslp t'/a|!lrt ewt to nhulpl tto q lno t4d$on.t .rn ptaqu .pJndtL 9 .mflll

1,,i,l;lll*i.,illr :; :;l

llil!!llllill!llli ri il

(q)

'sla8 puorsuaturpo rt cgoetord (llr) puB .si(v:rsoJcRucguolducsur4 (u) 'sle8 8urd.{l cruouaE (D ruorg purgdrcsaSeun crurouaS IEuoDcurU aJrlua uo etqBcllddB ,(lalll aJrI^WS We hryj aql 'sl tBql qods put stop spr?/yrol sprrBqtlo{ spefqo aturul-oq ;o anp^ uol&cfiddr pepmlxa Joaq ,(uur ru4si(s rply aqf, .sl4lg luarcgord pw uelaurerudI8urwu qll$ Jtrys prrB t4ry9 apq II uels,{s yJo sljedlueuoduoo uouBculjeds aql 'snqJ .sluaurE8{ tud,(toue8lurcgruSrs i(lJBcfotofq spr8.rot ptoqsaql "Juetdaccr OgeID Jo a8llu"cJad dlrsualur euEI Sunrunldo JoJ ,(lrsualu Og

arDI Ttopqrau8at puoe

'sJrBd eseq 6l.Z ol dn ]e uoq?r oppr8pusts Jo fcueda:csrp rofuu ar0 put spuEq Sulssgr pwslwod SqFoU qI^\ step rzrs lueur8s{ eql para^llap VCVSJo spJBpu4s azs vNC ueouu puB scuaunNorg Jo arBAttJos

.ru'rrowt&roq@tnD u o,

I

-------

r l t

t2

ll

t4

2 2 1

22

3 J r

32

ll9.E3

11&73

113,.16

112.45

vo 0,N

719 0.M

tt4 0.u)

tt! 0.tn

Table 1, Tmnslaled size dala bf Aida slttem and BioNumericssofrx)ate vith MarhQof and SAGA size standadt.

SL BO BioN SD BioN.M SD Aida SD

The authors would like to thank the National ScienceCouncil (NSC), the Council of Agriculrurc (COA), aDd rheDepartmenr of Indust al Technology (DoIT) of Ministryof Economic Aflairs (MOEA), Taiwan for financialsupports: NSC 94-2213-E-007-089, NSC 92 2745-B-034001, COA 93 AS-3.1.2-AD-U1, and MOEA 93-EC-17-A-l9,s 1-0016.

Refercnces

[1] Umeshadiga PS, Bhornra A, Turri MG Nicod A, DartaSR, Jeavons P, Mott R and Flint J (2c[l\. Autoraticanalysis of agarose gel irnager Bioinformatics 17:1084-1089.[2] Urneshadiga PS and Chaudhuri BB (2000).Segmentation and couating of FISH signals in confocalmicroscop! itages. Miqon 3l: 5-15.[3] Ye XY, Suen CY Cheriet M and Wang E (1999). ARecent Development in Inage Analysis of ElectrophoresisGels. Vsion Interface '99, Trois-Rivieres, Canada, 19 21.[5] ftao J, Shimazu Y, Ohra K, Hayasaka R andMatsushita Y (1996). Az outstandingness oriented irnageseqnEntation and its applkdrion. lnt. S)'rnposium onSignal Processing and its Applications, August.[6] Tsai MY. Chang CF, Chu HT, Chan CH, Chans KJ.Kao CY and Chen CC (2005). lt4icroarral liwgetPrc-Analysis for Critical Gene E prcision Computationwith Implemcnted Algo thmic Kemel, Hwa Kang Joumalof Agriculture 16: 43-50.[7] Appel RD, Vargas JR, Palagi PM, Walther D, andHochstrasser DF (1997). Melanie II - a third-peherationnftware package for analJsis ol two-dimznsionolelectrophoreris itnges: II. Algorithms. Electrophoresis l8:2135-2748.[8] Bettens E, Scheunden P, Sijbers J, Van DD, and MoensL (1996). Autonatic segmcntaion and modeting oftrroditvnsional electrcphorcsis gek. lntemationalConference on Image Processing 2: 665-668.[9] Umeshadiga PS and Chaudhuri BB (1999). Efficie tCell Segtvhtaion Tbol for Confocal Micmscop! TissueImages and Quahtitative Eyaluation of FISH Signals.Int. J. Microscopy Res. Technique 44r 49-68.I10l Umeshadiga PS and Chaudhuri BB (198).,4nelJicient cell segnentation tool for confocal microscopJtissue inuges fof quantitatire evaluation of FISH allelebards. lnt. J. Microscopy Res. Techniques 43: l-20.[1] Conradsen K and Pedersen ! (1992). Analysis oftt)o dimensional electrophoresis geh. Bionretrics 42:1273-128'7.It2] Lindeberg T (1998). Fedture Detection ,NithAutonatic Scale Selectioh. Intematlonal Joumal ofComputer Vision 30: 79-116.u3l McCarthy PJ, Sweetman SFS, McKenna PG andMcKelvey M (1997). Eroluation of manual and imageanalJsis quahtification of DNA tlamage in the alkalineconvt assat. Mlla,ger\esis 12: 209-214.

0.26

0,28

0.29

0.39

119.41 0.05

14.92 0.02

tt].Tt 0.N

112.12 0.00

11t.48

112.51

tt4 0.00

tt! 0.00

0.22

0.15

ll!.90 0.06

172.U 0.14

119.19 0.06

tu.t9 2.19

ta-2a 471.tE

113.E9 0.2tt

112.6 165.01

119,91 0.17

rr9,07 0.30

no 0,n179 0.u)

4 4 1

42

43

5 5 1

52

53

54

6 6 1

62

t23.02 0.42

D1.99 417

ll1.IB 0.47

1t2.99 0.19

123 0.d)

r22 0.00

tt4 01n

r13 0.M

122-E9

121.99

9.q)

ll9.m

t23 0.00

722 0.U)t20 0.m

719 0.u)

0.02

0.19

0,12

0.01

122, 0.19

nLal 0.19

119,99 0.r9

llg.to 0.16

720.U 0.21 120 0,00719.$ 0.29 719 0.(t0

L SU-6 and 8O an sir sanpb lanes with rcspe.ttue band objects onhplicate genotrping eel inases in lefa-nstu and apao-do|'n o er.2. RioN and BioN.M ttands lo, BioNunencs sofr|'amith IAGA andMatkQofr size stando^ lor data dDallsis in llodins poi ts thta..3. Aida is Aida qsten vith Ma*Qofr si.. ttaada s fu da,a a"atrsis irl

1 SD stattds Iot standttd deyiotion of iipli.de ahta sets.. The * indicatesab,ornal SD with nissine datd of BO in the AioN case.

The implemented cAM algoriihm imitates visualperc€ption in BO segnentation with peak and valley torefine BO ard to separate overlapped BO as well as withaverage band size to filter noise BO. Likely, microarray dotobJects (DO) with weak intensity amount demands belterimage analysis by GAM to filter noise DO towards besraccuracy by avoiding false-positive signal intensity whichmay be extrinsically introduc€d by noise object locatedwithin the weak dot object region.

Moreover, the implemented SAM algorithm maydemand custom-made approaches based on diiferent taskgoals of genomic, transcriptomic, arid proteomic analysis.The conmon theme is to defeat non-ideal image quality asof processing difficulty towards best data translation afterproficient grouping of autonatic analysis. Genotypic Aiddsystem achieves automatic lane finding by overcomingnon-st.aight lines of distorted and deformed images. Yet,commercial genotyping software defining lanes by manualdrawing or assigning rectangles realistically unlikely tolocate lanes and objects is less efficient and accurate. Verylikely, two-dimensional gel spot objects (SO) of conlpositernobility lanes demands better image analysis by SA,tl togroup and match geometric chessboard SO as of isoelectricfocusing and molecular weight separation perlormed inorthogonal directions iowards best accuracy by avoidingfalse-positive global superimposing which is mistakenlyaccomplishal by the erroneous feahrre SO selected for SOimage superimposing between experiment and control set.With accurate superimposing, the inter gel protein anrountcomparison of SO signal level on 2D-gel image set may befaithtully achieved for observing constituent and amountdifferenc€s between experiment and control samples.

5 Acknowledgments

EO.M

9.14

0.30

0.23

- 1328 -