1. PANTHER (Protein Analysis Through Evolutionary Relationships): Trees, Hidden Markov Models, Biological Annotations
Paul Thomas, Ph.D.
Division of Bioinformatics
Department of Preventive Medicine
Keck School of Medicine
University of Southern California

2. Outline
Brief history of PANTHER
What are the main features of PANTHER?
How are trees used in PANTHER?
What annotations can be obtained from PANTHER?
How do I use PANTHER in InterPro?

3. Brief History
Initial papers on PANTHER database and tools
New tree algorithm Thomas PD. BMC bioinformatics 2010
PANTHER 12 launched
- HMMER3
1998
2001
2003
2005
2008
2010
2013
2017
PANTHER joins InterPro
Integrated into GO Reference Genome curation pipeline.
2009, PLoS Comput. Biology
Gaudet P, 2011, Brief Bioinformatics
Start of the project
Human genome.
Use of UniProt reference proteomes

4. PANTHER features Comprehensive Covers eukaryotes and prokaryotes
Genome coverage depends on species
Vertebrates: 82% - 97%
Invertebrates: 50% - 80%
Fungi: 45% - 81%
Plants: 54% - 89%
Bacteria: 40% - 82%

5. Built on entire sequences, not single domains
PANTHER features
Built on entire sequences, not single domains
Designed to classify at two distinct levels
Family: homologous over at least part of the sequence
Subfamily: generally orthologous, i.e. descending from the same gene in common ancestral genome
Extensively annotated
Protein function (Gene Ontology)
Protein class (classifications “above” family level)

6. PANTHER Library Building
Curation
Annotated families and subfamilies
~1,300,000 sequences from 103 organisms
There are a number of tools available from PANTHER: I will focus on classification of protein sequences
Web services
~1,000,000 sequences
In ~15,000 family clusters

7. PANTHER Library Building
Curation
Annotated families and subfamilies
~1,300,000 sequences from 103 organisms
Web services
~1,000,000 sequences
In ~15,000 family clusters

8. PANTHER is a collaborative project
Curation
Annotated families and subfamilies
~1,300,000 sequences from 103 organisms
Web services
~1,000,000 sequences
In ~15,000 family clusters

9. Families, Trees and Subfamilies

10. Families, Trees and Subfamilies
duplication
Subfamily nodes
speciation

11. Annotations of HMMs Gene Ontology Gene/protein function
PANTHER “GO-slim”
Shows only selected terms from GO
Currently 550 terms vs ~45,000 for entire GO
Examples:
Molecular function: INSR + ‘receptor activity’
Cellular component: INSR + ‘plasma membrane’
Biological process: INSR + ‘insulin receptor signaling pathway’
Molecular level
Where molecule functions
What larger processes (biological programs) it is used in

12. GO Phylogenetic Annotation Project
Review experimental GO annotations
Build a model of function evolution
Gain and loss of specific functions along tree branches
Model predicts functions for unannotated genes
Gaudet, P., et al. (2011). Phylogenetic-based propagation of functional annotations within the Gene Ontology consortium. Briefings in Bioinformatics, 12(5), 449–62. doi: /bib/bbr042
PAINT
Phylogenetic Annotation and Inference Tool

13. carboxylic ester hydrolase
Green indicates experimental
Black dot indicates direct experimental data.
White dot indicates a more general functional class inferred from ontology
carboxylic ester hydrolase
Red indicates NOT function for the gene
cholinesterase

14. carboxylic ester hydrolase
Inheritance of functions through tree, unless following a loss
If a corresponding experimental annotation is not already present, these are PREDICTIONS
Node with loss of function
carboxylic ester hydrolase
Node with gain of function- cholinesterase
neuroligins
Gaudet, P., et al. (2011). Phylogenetic-based propagation of functional annotations within the Gene Ontology consortium. Briefings in Bioinformatics, 12(5), 449–62. doi: /bib/bbr042

15. Annotations of HMMs Protein class
~250 distinct terms arranged into a DAG
Groupings of protein families and subfamilies
Related functions
E.g. homeodomain vs. C2H2 zinc finger transcription factor
Enzymatic mechanism
E.g. cysteine protease vs. serine protease
Substrate specificity
E.g. potassium channel vs. sodium channel
Protein class groups based on diverse criteria that are based on classifications used by experts in a given area

16. PANTHER HMM scoring

17. PANTHER HMM scoring PANTHER Hit
-PANTHER family and subfamily (if HMM score > family)
-name (e.g., WNT-RELATED)

18. PANTHER family HMM annotation

19. PANTHER family HMM annotation

20. PANTHER family HMM annotation

21. PANTHER HMM scoring PANTHER Hit
-PANTHER family and subfamily (if HMM score > family)
-name (e.g., WNT-RELATED)

22. PANTHER subfamily HMM annotation

23. Summary PANTHER uses phylogenetic trees to classify protein sequences
Sequences in 103 genomes have already been classified at pantherdb.org
New sequences can be classified using PANTHER HMMs in InterProScan
Detailed annotation can be derived from PANTHER HMM hits
Gene Ontology
Protein Class
Pathways

24. Acknowledgements USC Huaiyu Mi Anushya Muruganujan Xiaosong Huang
John Casagrande
Sagar Poudel
GO Consortium
Suzanna Lewis
Pascale Gaudet
Marc Feuermann
Quest for Orthologs Consortium
Brigitte Boeckmann
Christophe Dessimoz
Adrian Altenhoff
UniProt (Reference Proteomes Project)
Maria Martin
Alan Wilter de Sousa
InterPro
Rob Finn
Lorna Richardson
Alex Mitchell
Neil Rawlings
pantherdb.org