2017 NIH SCAP-T: Data Analysis, Standards, and Benchmarks for Single Cell Analysis

(Stephen Fisher) #1

Wed, June 28th, Pre-Meeting Breakout Session

Moderator: Junhyong Kim (University of Pennsylvania)

Because of the difficulty of obtaining measurements at the single cell scale, the field has been driven by technological advances, including various RNA/DNA sequencing technologies, high-resolution proteomics and metabolomics, multiplexing strategies, cell handling technologies, etc. Despite these technological advances, single cell measurements remain difficult and is fundamentally challenged by the fact that the units of measurement, each cell, has no replication. It has been extremely difficult to assess the efficiency of measurements, establish benchmarks or controls, agree on protocols for data analysis, and coherently define standards for reporting experiments and data analysis. An especially important challenge is placing single cell data in their organismal context, including spatial coordinates.

Questions for this breakout session to consider include:

  • Is there benchmark data to compare new experimental or computational methods?
  • How do we establish material standards such as specific cells or spike-in RNA?
  • What metadata about calibration is important to know?
  • What information is important to collect about the sample and its preparation?
  • How can we work together with manufacturers to build standards into their methods?
  • Does an ontology need to be established for single cell analysis?
  • How can we associate single cells to tissue orientation information? More generally, how can data be organized from the single cell scale to whole organism scale?
  • What are the common data elements between imaging and sequencing assays? Is there a common header we can use for all data, similar to FITS or DICOM?

(Di Wu) #2

Different sequencing technologies generate different dropout rates. Shall the statistical methodology development of data analysis particularly take care this for different technology? Again, this may be related to the benchmark for evaluation. Similar samples sequenced by different technologies could be helpful too.

(Junhyong Kim) #3

This is a very good point. We probably need platform specific error models…also, better efforts to model those errors from empirical data.

(Terence Shine) #4

Excellent point!Those errors from empirical data drive me nuts.

(Stephen Fisher) #5

Here is a slide deck for the discussion later today.

HuBMAP2.pdf (1.7 MB)

(Hsinyi Tsang) #6

Is this even available to everyone attending the conference? Is there a room number at 6601 Executive Blvd? Thanks!

(Junhyong Kim) #7

The pre-Workshop is being held at Conference Room D in the Neuroscience Center (on the ground floor of 6001 Executive Blvd, Bethesda, MD).

(Zorina Galis) #8

YES! Please come on over!

Conference Room 105 - will be on the ground floor of the building – the posters and guard will direct you
Looking forward!

(Terence Shine) #9

Count me in! Count me in!