A Functional Genomics Approach to Autophagic Cell Death Gene Discovery
Gene Discovery by use of MySQL
description
Transcript of Gene Discovery by use of MySQL
Gene Discovery by use ofMySQL
• Background – myself
• NsGene – DTU satellite• Parkinson Disease (Affymetrix GeneChip)
• Analysis of fetal brain tissue• Search for new protein families
• MySQL & bioinformatic tools
Background• Thomas Nordahl Petersen
•Chemist, Ph.D protein Crystallography, University of Copenhagen
•Computational Scientist, SBI-AT (Hørsholm)
•Prediction of protein structure, secondary structure, fold recognition, homology modeling
•Bioinformatics - Gene discovery, NsGeneDevolop novel cell and gene based products for thetreatment of neurological diseases.
• Growth of cells in a capsule matrix
• The therapeutic protein be released directly in the relevant brain area
• Safe delivery across the blood-brain-barrier
ECT ProductsECT for Parkinson’s Disease
• Michael J. Fox foundation granted US $3 million to support a clinical “proof-of-concept” (May 2004)
• Identification of novel genes by use of bioinformatics
• NBN (GDNF family – potent neuroprotective effects)
Factor Products
• Scanning the human genome or assembled protein sets for different features of interest
A case study
• Search for Parkinson related gene(s)
• Affymetrix GeneChip experiments
• Fetal brain tissue
Parkinson DiseaseDegenerative central nervous system (CNS) disorder
Parkinson DiseaseLoss of dopamine producing brain cells
Parkinson’s Disease
• Dopamine from Substantia nigra activates neurons in Striatum/Basal ganglia
• Important for initiation of movement
Cure for Parkinson’s Disease ?
Parkinson disease may be cured provided that new dopamineproducing cells replace the dead ones.
Dopamin producing brain cells from aborted foetuses have beenoperated into the brain of parkinson patients and ín some casescured the disease. Brain tissue from approx 6 foetuses were needed.
Major ethical problems !
Search for a protein drug is the only valid option
Parkinson DiseaseDopamine producing cells
• Dopaminergic neurons can be found in the ventral part of the mesencephalon (VM) from approximately 6 weeks
• No dopaminergic neurons can be found in the neighbouring dorsal part (DM).
• Dopaminergic differentiation by use of GeneChips to compare the expression profiles of VM and DM
Fetal brain tissueMidbrain mesencephalon
Vm Dm+ Dopamine producingcells
- Dopamine producingcells
• Aborted feotus brain tissue – Karolinska hospital
• Feotus of age 6-10 weeks, 2 cases
Midbrain mesencephalon
Vm Dm
+ Dopamineproducingcells
- Dopamine producingcells
RNA purification + amplification
Affymetrix genechip analysis
Isolate the two samples (Vm/Dm)
Dopamine producing cells at the interface ?
GenePublisher(program by Steen Knudsen)
•Scale, normalize the Affymetrix GeneChip experiments
A1 A2 A2 B1 B2 B2 P-value
319 315 314 44 48 38 1.26e-07
314 334 327 443 434 444 6.55e-05
1980 1974 1973 1801 1785 1763 6.77e-05
123 123 126 87 88 93 8.01e-05
103 101 104 77 78 73 0.000112
107 107 111 79 77 82 0.000124
128 123 117 189 184 196 0.000142
179 179 186 145 147 149 0.000191
78 77 79 86 87 87 0.000202
96 90 93 136 129 138 0.000215
Vulcano plot
P-value
Log2 Fold change
Assigning Affymetrix GeneChip probes to a protein sequence
~20.000 probes on each of the A/B Affymetrix chips. Theprobes are normally not a part of a protein sequence.
Affymetrix probe
Blast
IPI protein sequence
Blast inferred
Unigene sequence (cDNA)5’ 3’
Internal database
Signal Peptide prediction
Conclusion – so far
• The most up-regulated genes include several ‘known’ genes like dopamine transporter (good positive control)
•The most interesting genes are the ‘unknowns’ that were up-regulated in Vm. Futher analysis is ongoing.
• Roland JR et al., Exp Neur (2006) Vol 198,2,427-437
“Identification of novel genes regulated in the developing human ventral mesencephalon”
A new growth factor family
• Criteria
• ‘Unknown’ family of protein sequences
• Growth factor like (Cys-Cys, SigP)
• Data source
• Assembled protein set/genomic data
• Search criteria are dynamic
• Use of MySQL
MySQL – a relational database language• Data are stored in tables as a ’black
box’
• Data physically separated from user
• Language is easy to read and understand
• Complex search queries
• Combine data in different tables/databases
• Result can be obtained in seconds
• Search criteria can be changed
Parsing Blast files(Preparing data for MySQL)
# Qname Dname Mlen Alen Qlen % a_id % q_id e-value Qfrom Qto Dlen Dfrom Dto
IPI00000001.1 STAU_HUMAN 577 577 577 100.0 100.0 0.0 1 577 577 1 577
IPI00000005.1 RASN_HUMAN 189 189 189 100.0 100.0 e-106 1 189 189 1 189
IPI00000006.1 RASH_HUMAN 189 189 189 100.0 100.0 e-106 1 189 189 1 189
IPI00000009.1 RASK_HUMAN 189 189 189 100.0 100.0 e-106 1 189 189 1 189
IPI00000010.1 RASL_HUMAN 188 188 188 100.0 100.0 e-105 1 188 188 1 188
IPI00000012.3 ZNT1_MOUSE 86 261 240 33.0 35.8 1e-32 1 230 503 248 500
IPI00000013.1 CSL2_HUMAN 334 334 334 100.0 100.0 0.0 1 334 334 1 334
IPI00000015.2 SFR4_HUMAN 494 494 494 100.0 100.0 0.0 1 494 494 1 494
IPI00000016.1 LMA3_MOUSE 114 145 145 78.6 78.6 9e-62 1 145 3333 1521 1665
Storing data from blast alignments
Field Type
query_db enum('hs_2_18','hs_2_23','affym','mm_1_11','affym_mouse')
query_acc varchar(20)
target_db enum('swissp','mm_1_11','sid','sid_mouse’)
target_acc varchar(20)
align_len smallint(6)
match_len smallint(6)
query_len smallint(6)
perc_align_len float(5,1)
perc_query_len float(5,1)
minus_ln_e float(6,2)
query_from smallint(6)
query_to smallint(6)
target_from smallint(6)
target_to smallint(6)
target_len int(11)
MySQl example
SELECT
a.query_db, a.query_acc,
a.target_db, a.target_acc,
a.perc_align_len, a.minus_ln_e,
b.target_db, b.target_acc,
c.cleavage_site
FROM
blastdb AS a, blastdb AS b, signalp AS c
WHERE
a.query_db='hs_2_23' AND a.target_db = 'mm_1_11' AND
a.target_acc != 'NULL' AND b.target_db='swissp' AND
a.query_acc=b.query_acc AND b.target_acc='NULL' AND
c.query_db='hs_2_23' AND c.query_acc = a.query_acc AND
c.cleavage_site >= 15 AND c.cleavage_site<=45;
Output from MySQL
query_db query_acc target_db target_acc perc_align_lenminus_ln_e target_db target_acc cleavage_site
hs_2_23 IPI00000111 mm_1_11 IPI00223686 48.6 999.00 swissp NULL 35
hs_2_23 IPI00000183 mm_1_11 IPI00108107 74.0 999.00 swissp NULL 26
hs_2_23 IPI00000381 mm_1_11 IPI00128682 78.5 206.13 swissp NULL 21
hs_2_23 IPI00001001 mm_1_11 IPI00221700 91.7 173.39 swissp NULL 45
hs_2_23 IPI00001443 mm_1_11 IPI00221913 60.0 17.73 swissp NULL 30
hs_2_23 IPI00001578 mm_1_11 IPI00122466 88.8 207.93 swissp NULL 38
hs_2_23 IPI00001719 mm_1_11 IPI00120961 83.1 52.27 swissp NULL 44
hs_2_23 IPI00001952 mm_1_11 IPI00225921 76.0 999.00 swissp NULL 44
hs_2_23 IPI00002173 mm_1_11 IPI00112960 85.4 999.00 swissp NULL 42
Clustering of protein sequencesTribe-mcl
47306 sequences
13130 clusters
Store in MySQL1) Cluster size
ACPGICSKSCCPFLTPALCSRTCCPY
2) Cys-Cys230
2
16
(3)
Conserved Cys-Cys
• Many growth factor families have their own specific Cys-pattern,TGF- family.
•Transforming growth factor- is a multifunctional peptide that controls proliferation, differentiation and other functions in many cell types.
• Search for Cys-pattern without any a priori knowledge
Search criteria
• Family cluster size > 1
• No SwissProt homologues
• Cys count > 4
• Signal Peptide
• Mouse homologue/orthologue
• 48 Families
• Manual inspection of alignments (- isoforms)
• Upload remaining sequences to internal database
Internal database
Tissue-specific expression10
0 bp
ladd
erU
nive
rsal
ref
Who
le b
rain
Hea
rtK
idne
yLi
ver
Lung
Plac
enta
Pr
osta
teSa
livar
y gl
and
Skel
etal
mus
cle
Sple
enTe
stis
100
bp la
dder
100
bp la
dder
Thym
usTh
yroi
d gl
and
Trac
hea
Ute
rus
Col
onSm
all I
ntes
tine
Spin
al C
ord
Feta
l Liv
erFe
tal b
rain
Panc
reas
Neu
rosp
here
ctrl
dH2O
100
bp la
dder
Outcome from Gene Search
• Family including 5 sequences
• At least 8 Cys
• Predicted as growth factors/hormones
• ~125 – 140 amino acidscys10A
cys10B
cys10C
cys10E
cys10D
Outcome from Gene Search• Family including 2 sequences - approx 30% seqid
• 11 of 16 Cys are conserved
• Effect on cultured neural cells