Open-Source Bioinformatics for Data Scientists with Amanda Schierz
- 1. Open Source Bioinformatics for Data Scientists Amanda Schierz
- 2. Recent Projects ! Druggability prediction ! 3D structure ! Protein Sequence ! Predict a protein’s druggability based on it’s position in the protein-protein interaction network ! Drug Resistance ! Therapeutic opportunities ! Identification of new gene targets for cancer ! Are they Druggable? ! Candidate Compounds ! Compounds more likely to be a hit for a bioassay
- 3. Drug Discovery Process Early-stage: Discovery Optimisation ADMET Clinical Trials Paperwork • Target Evaluation • Compound Screening • Computational Chemistry • Structure- based Drug Design • Absorption Distribution Metabolism Excretion Toxicity • Patient Stratification • Protocol • Drug Approval
- 4. Biology 101 ! There is a many to many relationship between Gene and Protein ! A Protein is a large molecule; a Drug is a small molecule ! Gene Expression data ! The amount of a gene produced. Epigenetics. ! highly / lowly / over / under – fold change ! Warning: Platforms and preprocessing ! Gene Copy Number ! Loss / Gain a gene ! On one strand or 2? ! There are only approx. 400 genetic targets of approved pharmaceuticals ! Only from a handful of Protein Families ! Desperate need for diversity
- 6. Target Identification ! Prediction of disease-associated genes ! patient level ! gene / protein level ! network ! Prediction of mechanisms of disease ! Epigenetic targets – meta-targets ! Prediction of protein function – from sequence / structure / network ! multi-class; multi-label ! Prediction of 3D structure ! Prediction of protein binding ! New immune targets
- 7. Druggability Prediction ! Drugs – FDA Approved ~350 Very strict – know therapeutic benefit ! Drugbank – loose – binds but no therapeutic benefit ! Tractable or Druggable ! Rule of 5 compliant ! Precedence-based - Druggable families / Homology - Ligand-based scoring - Uniprot, bioassays – EBI and Pubchem bioassay - Statistical analysis
- 8. Druggability Prediction ! Sequence Analysis - Amino Acid motifs and composition - Physicochemical descriptors - infinite amount – very wide data set - Supervised classification ! FASTA - can download all human sequences from Uniprot >seq0 FQTWEEFSRAAEKLYLADPMKVRVVLKYRHVDGNLCIKVTD ! R ProtR ; R Bioconductor ! species,mhc,peptide_length,A,R,N,D,C,E,Q,G,H,I,L,K,M,F,P,S,T,W,Y ,V,scl1.lag1,scl2.lag1,scl1.lag2,scl2.lag2,scl1.2.lag1,scl2.1.lag1,scl1.2.l ag2,scl2.1.lag2,AA,RA,NA,DA,CA,EA,QA,GA,HA,IA,LA,KA,MA,F A,PA,SA,TA,WA,YA,VA,AR,RR,NR,DR,CR ..... ,Schneider.Xr.K,Schn eider.Xr.M,Schneider.Xr.F, Grantham.Xr.A,Grantham.Xr.R,
- 9. Druggability Prediction ! 3D structure - Pockets, surface area - Ligand interaction fingerprints - Supervised classification
- 10. 3D Structure ! PDB, ProtDCal, PockDrug
- 11. Druggability Prediction ! Interaction Network ! Many use cases ! Data from EBI and Y2H ! List of binary interactions ! Becareful 1: Data is inherently biased ! Becareful 2: Complex interactions ! R iGraph; Gephi for visualisation ! Topological properties ! Community analysis ! Subgraph analysis ! Statistical analysis, network analysis and supervised classification
- 13. Drug Resistance
- 14. Drug Resistance
- 15. Compound Bioactivity ! Brute force mass screening ! 1000s compounds screened in batches ! Primary Assays; Secondary / confirmatory assays ! Can be binary classification or regression ! The IC50 is a measure of how effective a drug is. ! Active / inactive : IC50 threshold ! Goal is also to identify diverse compound structures ! Scaffold Hopping ! Same kind of method as Protein Sequence conversion ! Pharmacophore fingerprints ! https://www.chemaxon.com/free-software/
- 17. Compound ADMET ! Many use cases ! ADMET of hits ! Absorption ! Distribution ! Metabolism ! Excretion ! Toxicity ! Mutagenecity ! Protein binding
- 18. General Resources ! EBI European Bioinformatics Institute / Pubchem ! API ! Integrates several downloadable Data Sources (expression, Copy Number, Bioassays, network, disease-specific) ! Baseline data (Normal not diseased) ! Protein Data Bank – 3D Structures ! DrugBank ! Cancer – The Cancer Genome Atlas (TCGA) and International Cancer Genome Consortium (ICGC) ! Coding Tools – R Bioconductor , BioPerl, BioPython ! https://docs.chemaxon.com/display/docs/Documentation
- 19. General Resources ! canSAR database ! Integration of biological, pharmacological, chemical, structural biology and protein network data
- 20. Beware 101 ! Non-standard Gene names ! Some experiments Genes, some are Proteins ! We need new Drug Targets, different from established ones. ! Keep in mind when analysing results ! Cancer is difficult ! Drug resistance ! Data is not up with the science ! Tumour Heterogeneity ! Wide data = random patterns ! Different expression / sequencing platforms
- 21. Therapeutic Opportunities ! Approximately only 350 - 400 protein targets ! DNA damage response (DDR) is essential for maintaining the genomic integrity of the cell ! Currently targeted by chemotherapy and radiation. Goal is for small molecule targeting ! TCGA Patient Analysis: Expression, Copy Number Variation and Mutation data. ! 15 cancer disease types ! Telegraph March 2015 ! New drugs to tackle cancer cell weak spots could end 'scattergun' chemotherapy Laurence H. Pearl, Amanda C. Schierz, Simon E. Ward, Bissan Al-Lazikani, Frances M. G. Pearl. Therapeutic opportunities within the DNA Damage Response. Nature Cancer Reviews
- 22. Therapeutic Opportunities ! Statistical analysis of DDR deregulation in patients compared to a random set of genes ! Druggability prediction of deregulated DDR genes ! Synthetic Lethality analysis of Yeast DDR orthologues ! Two genes are synthetic lethal if mutation of either alone is fine but mutation of both leads to cell death. Targeting a gene that is synthetic lethal to a cancer-relevant mutation theoretically will kill only cancer cells.