Biological Databases

1
Biological Databases

• Asif Jan
• Brain Mind Institute
• EPFL

2
Introduction

• The purpose of the biological experiment is to
  understand working of the biological organs such
  as brain, cells etc
• The purpose is to study the interplay of
  different structural, chemical and electrical
  signals that gave rise to natural and disease
  processes.
• The data is acquired from different angles to
  serve different research purposes i.e. different
  animal models, using physiological approaches,
  anatomical approach and levels of protein
  activities etc

3
Sequence

• Biological Data - Types and Constraints
• Neocortical Microcircuit Data
• Databasing Neocortical Microcircuit
• Conclusion and Discussion

4
Part I

• Biological Databases

5
Biological Databases- Challenges

• A great deal of diversity in the data types
• Unconventional and adhoc query requirements
• Ubiquitous uncertainty in the data
• Requirements for data curation
• Need for detailed Data annotations
• A need for large scale data integration
• Non-availability of universal taxonomy
• Support for rapid Schema evolution
• Temporal Data Management
• (Ref Data Management for Molecular and Cell
  Biology - Workshop report www.lbl.gov/olken/wdmbi
  o )

6
Data Types (1/2)

• Sequences
• DNA, RNA, amino-acid sequences (proteins). The
  data has grown enormously due to availability of
  automated sequence machines and large scale
  sequencing projects such as human and mouse
  genome.
• Graphs
• Biological pathways such as metabolic pathways,
  gene regulatory networks, 3D protein structures
• High Dimensional Data
• Micro-array experiments (thousands of genes),
  hundreds of experimental conditions, clustering
  studies on genes etc
• Shapes
• 3D molecular structural data augmented by
  chemical distribution, 3D cell morphology data

7
Data Types (2/2)

• Temporal Data
• Useful for studying the dynamics of biological
  system e.g. electrophysiology recordings,
  development biology, protein structure dynamics,
  cellular structure dynamics etc.
• Model Data
• Representation of biological phenomenon as
  computational, mathematical and statistical
  models used for parameter estimation, testing
  etc. Models shall also be represented and stored
  in query-able format.
• Scalar and Vector Fields
• Charge distribution across cell surface, calcium
  and protein fluxes across cell surface etc
• Extracted Features Data
• Numerical data extracted from the combination of
  one of the above data types

8
Adhoc Query Requirements

• Biologists understand the relation across
  different data types and these relations are not
  necessary obvious from the database point of view
  i.e.
• Two labs one studying dendritic spines of PC
  in hippocampus, primary schema element being the
  anatomical entities (dendrites etc) reconstructed
  from 3D serial sections. The other studying
  Purkinje cells in the cerebellum branching
  patterns from the dendrites of neurons and
  protein localization in various compartments
• Thus a researcher, modeling effects of
  neurotransmission in hippocampal spines would get
  structural information from lab 1 and information
  on calcium binding proteins found in spines from
  lab 2. Assumption Like PC, Purkinje cells also
  possess dendritic spines and release of calcium
  in spiny dendrites occurs as a result of
  neurotransmission and causes change in spine
  morphology. Propagation of calcium signals
  through out the neuron depends on the morphology
  of dendrites
• Ref Ludascher B, Gupta A and Martone M, Model
  Based Mediator System for Scientific Data
  Management

9
Uncertainty in the data

• Biological data has a great deal of uncertainty
  as it represents a biological phenomenon that is
  observed and assumed (based on some evidence) to
  be true.
• For example, the spiking behavior of cell under
  specific stimuli, protein sequence in the protein
  database that is based on partial protein report
  etc.
• The uncertainty must also be modeled and recorded
  as part of the data as it has consequences for
  subsequent usage of the data.

10
Requirements for data curation

• The data is collected across different structural
  and function boundaries, there might be many
  missing links and inconsistencies (some
  inconsistencies due to lack of core domain
  knowledge etc).
• Often is the case that expert intervention is
  required for cross correlation of the data and
  for filling in missing links and/or improving the
  data consistency.
• However, large scale biological database entail
  explicit representation of uncertainties and
  cross structural/functional boundaries in order
  to have automatic curation.

11
Data Annotations

• Biological data is specific to the purpose of the
  individual performing the data collection.
• For example, while studying calcium regulation
  researcher A might adapt a physiological approach
  using patch electrodes and researcher B may take
  anatomical approach mapping different isoforms of
  calcium current to structure of organelles that
  expresses them etc.
• Different animal models, need to integrate data
  collected from different brain regions, across
  different species etc. Furthermore, the
  experimental conditions have a great influence on
  the experimental results.
• This requires that the data shall be properly
  annotated during different stages of data
  collection and all conditions/parameters properly
  recorded. Furthermore, assumptions in doing
  experiment etc shall also be recorded.
• In case of data derived from primary data the
  need for proper annotations is further enhanced.
  

12
Need for large scale data integration

• It is very difficult, if not impossible, to
  collect information about various biological
  entities at a single institute or laboratory.
• Data collected from years of research, across
  different functional and anatomical scales, and
  for normal as well as disease cases is available
  for use.
• Often this data is poorly annotated and
  inadequately structured yet contains precious
  information that can not be ignored.
• However, lack of universal taxonomy, or a uniform
  structure presents many challenges for effective
  utilization of this data.
• While improving the readability of the existing
  databases, it is imperative for the new databases
  to adapt proper descriptions, query interfaces
  and annotation frameworks.

13
Other Issues

• Lack of Taxonomies
• Schema Evolution
• Biological Constraints
• Data Cleaning

14
Part II

• Neocortical Microcircuit Data

15
The Neocortex and the Cortical Column
Cortical Sheet
Neocortex
Cortical Column
Layer I II III
IV V VI
16

• Key properties of neurons and synapses
  numerically represented as profiles

17
Neuron ProfilesMorphology Data (m-Profile)

• Neuron 3D reconstructed and converted to
  Neurolucida format
• Analyzed by a MATLAB based tool to extract a
  vector of 200 values
• Values can be used to artificially rebuild
  neurons with specific statistical properties
• Example Parameters
• TreeLengthMean (mean of lengths of segments with
  same order in each tree)
• IndivTreeLengthMean (mean of segment lengths in a
  tree)
• XY_Angle (angle between projection of a segment
  on XY plane and X axis)

18
(No Transcript)
19
Neuron ProfilesElectrophysiology Data (e-Profile)

• Obtained by applying a series of current
  injections to the samota
• Response measured to obtain a spectrum ( 140) of
  electrophysiological parameters (EPs)
• Most parameters sensitive to the ion channel
  composition
• Raw electrophysiological data also stored
• Example parameters
• ADP (after depolarization immediately following
  APs)
• APThreshhold (threshold of AP generation during a
  ramp polarization)
• SineSpectrum (various measures of frequency
  filtering by neuron)
• sAHP (amplitude of hyperpolarization after a
  burst of APs)

20
(No Transcript)
21
Neuron ProfilesGene Expression Data (g-Profile)

• Obtained from single cell multiplex RT-PCR
  studies and single cell DNA microarray analyses
• Enable non quantitative detection of expression
  vs non expression of 50 genes
• Extended the system to conduct single cell DNA
  microarray studies to screen for over 20,000
  genes

22
(No Transcript)
23
Synaptic ProfilesMorphology Data (sm-Profile)

• Describes anatomy of a synaptic connection
• Contains information about number of synapses,
  their location on axonal and dendritic arbors of
  pre , post synaptic neuron, axonal and dendritic
  geometric and electronic distances
• Examples
• Axonal Branch Order (number of branch points
  between the bouton forming the synapse and the
  soma of the source neuron)
• Dendritic Branch Order (the location of the
  synapse along the dendritic arbor according to
  the branching frequency of the dendritic tree)
• Geometrical Distance (the distance along the
  dendritic from the synaptic location to the
  postsynaptic soma)

24
(No Transcript)
25
Synaptic ProfilesElectrophysiological Data
(se-Profile)

• Characterized in terms of the biophysical
  dynamic properties
• The biophysical properties focus on the
  amplitudes, latencies, rise and decay times
  synaptic conductances synaptic charge transfer,
  etc
• The dynamic properties include the time-constants
  governing the rates of recovery from synaptic
  depression (D) and facilitation (F) as well as
  the absolute and effective utilization of
  synaptic efficacy parameters
• Other parameters include estimates of probability
  of release and number of functional release sites

26
(No Transcript)
27
Synaptic ProfilesPharmacological Data
(sp-Profile)

• Contains information describing the sensitivity
  of the synaptic connection to different chemicals
• Described in terms of synapse response to various
  blockers, agonists and antagonists
• Commonly used chemicals are
• bicuculine (GABA-a antagonist)
• APV (NMDA receptor antagonist)
• CNQX (AMPA receptor antagonist)
• CGP 35348 (GABA-b antagonist)
• NMDA (NMDA receptor agonist)
• diazepam (GABA-a facilitator)

28
Additional Data

• General Data (g-Profile)
• Animal Information
• Brain Region Information
• Experiment Information
• Model Data (mod-Profile)
• A complete NEURON model that will include
• active properties by inclusion of ion channel
  constellations and parameters
• electrical properties of the neurons
• possible ion channel constellations
• Canonical Data (x-Profile)
• statistical analysis of stored neurons and
  synapses
• these simplified models to be used for
  visualization and simulation etc

29
Part III

• Databasing Neocortical Microcircuit

30
Blue Brain Project - Goals

• Gather and share raw data about different facets
  of the neurons
• Develop biologically accurate model of neurons
  and their interactions
• Obtain a biologically accurate simulation of the
  cortical column
• Develop visualization tools on the simulation to
  perform in silico biology studies

31
Blue Brain Data Usage
Visualization
Produces data
Simulation
Are used for
Experiments
Models
Test against Experimental data
Is stored
Produces statistics On experiments
Database
32
Overview of current system
33
Storage Resource Broker

• Distributed data storage
• Uniform access to a variety of data sources
• Intuitive file system-like interface with the
  data
• Ability to annotate data with metadata
• Data access management
• Security management

34
SRB Federation architecture
35
SRB Server Architecture
36
The Metadata Catalog (MCAT)

• Stores and manages the information about an
• SRB System
• Physical and connection information about the
  data sources
• User information and privileges
• Logical and physical mapping to the files
• Files metadata
• System metadata
• User-defined metadata

37
Current Database Infrastructure

• Uses SRB for storing the primary data
• Customized tools for data upload, annotation and
  download
• Metadata about primary data is stored in the
  metadata catalog
• Secondary databases storing morphology,
  electrophysiology and gene expression data
• Specialized database for storing extracted
  features

38
Current Set of Databases

• Morphology Data - structure of neurons
• Electrophysiology Data- Recordings from patch
  clamp system, response of neurons to stimuli
• Gene Expression Data expression of specific
  genes. Started to collect DNA microarrays
• Feature DB primary features extracted from
  electrophysiology data
• Synapse DB properties of synaptic connections,
  release probabilities , conductances etc
• Index DB A central bookkeeping and
  synchronization system.
• Microcircuit DB a description of the
  microcircuit with 10,000 neurons and 2 million
  connections

39
Morph DB
Synapse DB
EP DB
Gene Expr DB
Index DB
Feature DB
SRB System
Files
Relational DB
40
Part IV

• Conclusion and Discussions

41
Reapplying challenges

• Diversity - morph , electro, gene , statistics ,
  models
• Unconventional Queries - need for expert
  knowledge to draw relations
• Uncertainty in the data
• Need for annotations, taxonomy , metadata (person
  doing experiemnts on PC does not record the type
  of cell in his lab book)
• Schema Evolution
• Need for integration

42
Conclusion

• A basic framework for storing neocortical
  microcircuit data (primary) as well as metadata
• Tools for data upload, consistency checking, data
  download and browsing
• Capability to store different type of data i.e.
  images, recordings, ascii files etc
• Hierarchical database infrastructure supporting
  secondary and specialized database
• Flexible structure catering for new data types

43
Conclusion (in progress)

• Using standard taxonomy , Using ontologies to
  facilitate integration within various databases,
  and with external databases
• Data Integration across multiple databases for
  supporting experimentalists
• Development of efficient and user friendly
  interaction environments
• Knowledge based Query Environment
• Scalability Issues

44
Thank you