1. Genomic Data Manipulation
Curtis Huttenhower
TA: Fah Sathirapongsasuti
Office hours: R 2:00-3:00
MW 3:30-5:20
Office hours: F 3:30-5:20
Harvard School of Public Health
Department of Biostatistics

2. Genomic Data Manipulation
1/3 methods:
Quantitative methods (mini-stats)
Programming (Python)
1/3 applications:
DNA sequence data
Microarrays
Proteomics and metabolomics
Interaction networks
1/3 papers and projects:
Journal club
Final project
10% participation
50% problem sets (10)
15% presentations
25% final project

3. What tools enable biological discoveries?
Our job is to create computational microscopes:
To ask and answer specific biomedical questions using millions of experimental results

4. A computational definition of functional genomics
Prior knowledge
Genomic data
Gene ↓
Function
Gene ↓
Data ↓
Function
Function ↓

5. A framework for functional genomics
100Ms gene pairs → G1 G2 + G4 G9 … G3 G6 - G7 G8 G5 ?
0.9
0.7
0.1
0.2
0.8
0.5
0.05
0.6
← 1Ks datasets
P(G2-G5|Data) = 0.85
High
Correlation
Low
Frequency
High
Correlation
Low
Let.
Not let.
Frequency + =
Similar
Dissim.
Frequency
High
Similarity
Low

6. Functional network prediction and analysis
Global interaction network
Carbon metabolism network
Extracellular signaling network
Gut community network

7. Predicting gene function
Predicted relationships between genes
High
Confidence
Low
Cell cycle genes

8. Predicting gene function
Predicted relationships between genes
High
Confidence
Low
Cell cycle genes

9. Predicting gene function
Predicted relationships between genes
High
Confidence
Low
These edges provide a measure of how likely a gene is to specifically participate in the process of interest.
Cell cycle genes

10. Comprehensive validation of computational predictions
With David Hess, Amy Caudy
Genomic data
Prior knowledge
Computational Predictions of Gene Function
SPELL
Hibbs et al 2007
bioPIXIE
Myers et al 2005
MEFIT
Retraining
Genes predicted to function in mitochondrion organization and biogenesis
New known functions for correctly predicted genes
Could go (-)
Laboratory Experiments
Petite
frequency
Growth
curves
Confocal microscopy

11. Evaluating the performance of computational predictions
Genes involved in mitochondrion organization and biogenesis
106
Original GO Annotations
135
Under-annotations 82
Novel Confirmations,
First Iteration 17
Novel Confirmations,
Second Iteration
340 total: >3x previously known genes in ~5 person-months
Could go (-)

12. Evaluating the performance of computational predictions
Genes involved in mitochondrion organization and biogenesis
Computational predictions from large collections of genomic data can be accurate despite incomplete or misleading gold standards, and they continue to improve as additional data are incorporated.
106
Original GO Annotations 95
Under-annotations 40
Confirmed
Under-annotations 80
Novel Confirmations
First Iteration 17
Novel Confirmations
Second Iteration
340 total: >3x previously known genes in ~5 person-months
Could go (-)

14. Fig. 3. Comparison of Sargasso Sea scaffolds to Crenarchaeal clone 4B7.
Comparison of Sargasso Sea scaffolds to Crenarchaeal clone 4B7. Predicted proteins from 4B7 and the scaffolds showing significant homology to 4B7 by tBLASTx are arrayed in positional order along the x and y axes. Colored boxes represent BLASTp matches scoring at least 25% similarity and with an e value of better than 1e-5. Black vertical and horizontal lines delineate scaffold borders.
J C Venter et al. Science 2004;304:66-74
Published by AAAS

16. Fig. 7. Phylogenetic tree of rhodopsinlike genes in the Sargasso Sea data along with all homologs of these genes in GenBank.
Phylogenetic tree of rhodopsinlike genes in the Sargasso Sea data along with all homologs of these genes in GenBank. The sequences are colored according to the type of sample in which they were found: blue, cultured species; yellow, sequences from uncultured organisms in other environmental samples; and red, sequences from uncultured species in the Sargasso Sea. The tree was divided into what we propose are distinct subfamilies of sequences, which are labeled on the right. The tree was constructed as follows: (i) All homologs of halorhodopsin were identified in the predicted proteins from the Sargasso Sea assemblies using BLASTp searches with representatives of previously identified halorhodpsinlike protein families as query sequences. (ii) All sequences greater than 75 amino acids in length were aligned to each other using CLUSTALw, and a neighbor-joining phylogenetic tree was inferred using the protdist and neighbor programs of Phylip.
J C Venter et al. Science 2004;304:66-74
Published by AAAS

17. Aerobic, microaerobic and anaerobic communities

18. Model of microbial biomarkers