Regulator Discovery from Gene Expression Time Series of Malaria Parasites: a Hierarchical Approach
José Miguel HernándezLobato, Tjeerd Dijkstra and Tom Heskes
Universidad Autónoma de Madrid, Leiden University Medical Center, Radboud University Nijmegen

Poster ID T51

Abstract
We  introduce  a  hierarchical  Bayesian  model  for  the  discovery  of  regulators  from  gene  expression  data  only. The hierarchy incorporates the  knowledge that only a  few  regulators  regulate  most  of  the  genes.  Running  the  model  on  a  malaria  parasite  data  set,  we  found  four  genes  with  significant  homology  to  transcription  factors  in  an  amoebe,  one  RNA  regulator  and  three  genes  of  unknown  function (out of the top ten genes considered).

Method
We  approximate  the  posterior  distribution  of  the  model  parameters by means of expectation propagation.

R e s u lt s f o r th e 3 D 7 d a t a s e t f r o m Pla s m o d i u m .

Ra nk 1 2
...

Name PFC0950c PF11_0321
...

Annotatio n or BLASTP hits 25% Identity to GATA TF in Dictyostelium 25% Identity to WRKY TF in Dictyostelium
...

Experiments
We tested the algorithm on the cdc15 dataset from yeast  and  on  the  3D7,  Dd2  and  HB3  datasets  from  Plamodium.  We  took  the  10  genes  with  highest  probability  of  being  regulators  and  looked  up  their  annotations  and  sequence  homologies  at  public  databases: SGD and PlasmoDB.

5 6
...

PFD0175c

32% Identity to TF in Dictyostelium

MAL7P1.34 35% Identity to TF in Dictyostelium
... ...

The model
The expression of each gene is linearly regressed against  the expression of all other genes.  A hierarchical prior for  the  regression  coefficients  enforces  sparsity.  The  posterior  of  a  latent  parameter  determines  the  probability  of each gene being a regulator.

10

MAL13P1.14 DEAD box he licase

Results
Results  for  yeast  are  statistically  significant.  Results  for  3D7  and  HB3  are  also  satisfactory.  However,  only  one  regulator was found in the Dd2 dataset.
R e s u lt s f o r th e cd c 1 5 d a t a s e t from y e a s t .

T h e m o d e l a llo w s to d e t e r m in e th e g e n e s m o r e lik e ly to be reg u la t e d by a reg u la t o r .

Ra nk 1 2
...

Name

Anno tatio n

YLR098c DNA binding tra nscriptio na l activator YOR315w Puta tive tra nscriptio n factor
... ...

6
...

YLR095w Transcriptio n elo ngatio n
... ... ... ...

10