Measuring the Microbiome:Methods & Tools

class: center, middle, inverse, title-slide

# Measuring the Microbiome:Methods & Tools
## 📚EPID 674📚
### Brendan J. Kelly, MD, MS
### Kyle Bittinger, PhD
### Updated: 25 May 2020

---

background-image: url(data:image/png;base64,#svg/bacteria.svg)
background-size: 500px
background-position: 50% 50%
class: center, middle, inverse

# Welcome!

---

# Where am I? How did I get here?

.pull-left[

__Epidemiology 674__ 
 
 __Measuring the Microbiome:__

__Methods & Tools__

]

.pull-right[

You study epidemiology...  
.pad-left[
or microbiology...

or informatics...

or statistics*
]
]

.footnote[
*Great opportunity to learn from each other, but this will be challenging online.
]

---

# Course Goals

.pad-left[
- How do we measure the microbiome?

- How are microbiome data structured?

- What are the pitfalls and dangers of microbiome data analysis?

- What analytic methods and model frameworks help address those challenges?

- __Reproducibility and rigor in microbiome research (e.g., reproducible reports)__
  
]

---
background-image: url(data:image/png;base64,#svg/petri_mixed.svg)
background-size: 500px
background-position: 50% 50%
class: center, middle, inverse

# How to measure the microbiome?

---

# "The Microbiome"

.pad-left[

- Loose definition: “the microorganisms in a particular environment” (synonymous with “microbiota”)
  
- Strict definition: “the combined genetic material of the microorganisms in a particular environment”

- emphasis on nucleic acid sequencing 
    
    - “omics” often defined by the tools on which they depend

]

---

# Biology/Ecology + Technology

.pad-left[
- Biology/Ecology:

- previously undetected microbial species & interspecific interactions
    
    - clinical relevance of these species and interactions

- Technology:

- [depends on high-throughput ("next generation") sequencing](https://youtu.be/fCd6B5HRaZ8)

- __NOT__ “unbiased” surveys of microbial communities
    
    - sequencing error, wet and dry lab decisions...

]

---

# Wet Lab Decisions

.pad-left[
- Specimen acquisition:

- how are samples obtained?
    
    - how are samples processed? (e.g., cell-associated bacteria from BAL; difficult-to-lyse Mycobacteria or spores)

- Sequencing library preparation:

- amplicon sequencing?
    
    - metagenomic sequencing?

]

---

# Dry Lab Decisions

.pad-left[

- Sequence modality: short reads versus long reads?

- Barcodes and de-multiplexing: Golay codes -- how much error is too much error?

- Paired-end joining: how much overlap and how much error?

- Sequence clustering:

- amplicon: operational taxonomic units or sequence variants?
    
    - metagenomic: assignment vs assembly?

]

---
background-image: url(data:image/png;base64,#svg/spreadsheet_cell.svg)
background-size: 500px
background-position: 50% 50%
class: center, middle, inverse

# How are microbiome data structured?

---

# More Data, More Problems

.pad-left[

- Measurement methods and tools differ

- Data embed wet and dry lab choices

- __BUT__ data structures similar and create common concerns:

- sparse count data
    
    - misclassification risks
    
    - high-dimensional data (p >> n)
    
]

---

# Microbiome Data Structure

.pad-left[
- Sparse count data generated by sequence binning/assignment

- Typically 1e2 - 1e4 microbiome variables (exceed observations)

- Cross-sectional or time-series measures

- Always think on two levels:

- What's there? __generative model for observed microbiota / genes__

- How do we know? __measurement model given constraints of our tools__
  
]

---
background-image: url(data:image/png;base64,#svg/danger-svgrepo-com.svg)
background-size: 500px
background-position: 50% 35%
class: center, middle, inverse

# Pitfalls and dangers...

---

# How do microbiome studies go wrong?

.pad-left[
- "Big data" (p >> n) problems:

- BSTA 785: Statistical Methods for Genomic Data Analysis
    
    - BSTA 789: Big Data
    
    - we will briefly survey commonly used techniques (dimension reduction, supervised and unsupervised clustering strategies) to bolster study design and planning

- Microbiome-specific problems...

]

---

# How do microbiome studies go wrong?

.pad-left[
- What is adequate sampling to estimate composition? ("shallow shotgun")

- How are microbes shared across study participants? ("cage effects")

- How to link sequence data to extant knowledge? (binning ITS data)

- Absolute abundance versus relative abundance? (Gloor et al)

- How to discriminate sub-species community members? ("strain-level")

]

---
background-image: url(data:image/png;base64,#svg/stacked_barchart.svg)
background-size: 500px
background-position: 50% 50%
class: center, middle, inverse

# ... analysis methods and models

---

# Microbiome Analysis Methods

.pad-left[
- Summary statistics (alpha diversity)

- Distance metrics (beta diversity)

- Mixture models (Dirichlet multinomial mixtures)

- Non-parametric tests (PERMANOVA/adonis)

- Data transformations (centered log-ratio transform)

- __No best method to address all questions!__

]

---
background-image: url(data:image/png;base64,#svg/computer_talk.svg)
background-size: 500px
background-position: 50% 50%
class: center, middle, inverse

# How are we going to do this online?

---

# EPID 674 Approach

.pad-left[

- To understand these decisions, you need to make them…  
       … and then make them again another way...
       
- You can’t learn microbiome data analysis without looking at microbiome data

- You can’t learn R without looking at R code

- Microbiome data demand reproducible research strategies

]

---

# EPID 674 2020 (COVID-19 edition)

.pad-left[
- Class schedule (adjusted for online learning)

- Assignments and evaluations

- Outline of topics covered

- [rstudio.cloud](https://rstudio.cloud) (R & Rstudio) to __learn by doing!__

- Recommended resources

]

---

# EPID 674: Class Schedule

.pad-left[

- Classes Tuesdays & Thursdays May 26 - June 30, 2020

- Each class session:

1. lecture chunks (prerecorded for your convenience)
    
    2. readings & Canvas discussion
    
    3. R coding primers
    
    4. live discussion 3-4pm on BlueJeans: [https://bluejeans.com/9715046666](https://bluejeans.com/9715046666)
    
- Complete lectures, readings, & primers __before__ live sessions -- bring questions!

]

---

# EPID 674: Assignments

.pad-left[

- Assignments:

1. readings & Canvas discussion boards
    
    2. R code exercises
    
    3. in-class presentations
    
    4. final presentation
    (reproducible analysis report with RMarkdown)
    
]

---

# EPID 674: Evaluation

.pad-left[

- Grades:

- 40% participation
      (Canvas & BlueJeans discussion,
      in-class coding exercises,
      in-class presentations)
    
    - 30% homework (independent R code exercises)
    
    - 30% final presentation  
      (reproducible analysis report with Rmarkdown)

]

---

# EPID 674: Outline

.pad-left[

- Week 1: sequencing, sequence clusters & microbiome data

- Week 2: taxonomy, phylogeny, binning & (mis)classification

- Week 3: "diversity" as summary statistic or distance

- Week 4: compositional data analysis & mixture models

- Week 5: dimension reduction methods, ordination

- Week 6: planning microbiome studies

]

---

# EPID 674: Resources

.pad-left[

- [rstudio.cloud](https://rstudio.cloud)

- supported method for assignments & in-class exercises
    
    - if you want to install R & RStudio locally, please do so!  
    
      (R: download & install from [CRAN](https://cran.r-project.org/))  
      
      (Rstudio: download & install from [rstudio.com](https://www.rstudio.com))

]

---

# EPID 674: Resources

.pad-left[

- R programming & “tidyverse”:

- Wickham’s _R for Data Science_  
      ([https://r4ds.had.co.nz/](https://r4ds.had.co.nz/))

- Analysis:

- Legendre & Legendre’s _Numerical Ecology_
    
    - James, Witten, Hastie & Tibshirani’s  
      _Introduction to Statistical Learning with Applications in R_  
      ([http://faculty.marshall.usc.edu/gareth-james/ISL/](http://faculty.marshall.usc.edu/gareth-james/ISL/))
      
]

---
class: center, middle, inverse
background-image: url(data:image/png;base64,#svg/conjugation.svg)
background-size: 500px
background-position: 50% 50%

# Questions?
### Post to the discussion board!

---
background-image: url(data:image/png;base64,#svg/bacteria.svg)
background-size: 100px
background-position: 98% 90%
class: center, middle

# Thank you!
#### Slides available: [github.com/bjklab](https://www.gihub.com/bjklab/microbiome-measures_01_data-and-danger)
#### [brendank@pennmedicine.upenn.edu](brendank@pennmedicine.upenn.edu)