Notes 20101105 Heather - Meiering Group Meeting

From SnOwy - Ed's Wiki Notebook

Heather's giving a group meeting today about aggregation prediction programs and how they apply to SOD (superoxide dismutase)

Contents 1 Overview 2 Motivation 3 Sequence independent 3.1 Chiti-Dobson 3.2 Agar Method 3.3 Zyggregator 3.4 TANGO 3.5 PASTA 3.6 FoldAmyloid 4 Including structural information 4.1 WALTZ 4.2 ZipperDB 5 Predicting SOD Aggregation

Overview

1. sequence independent methods
2. sequence dependent methods
3. structure dependent methods

Motivation

• creating predictive models is a challenge -- we don't really understand everything involved in aggregation phenomenon
• aggregation pathways -- fibrils -- amorphous fibres etc
• predicting what will aggregate, as well as what the aggregates might look like (the inclusion body composition?)
• suggest unfolded items tend to create aggregates
• framework based on the thinking
• non-homogeneous system leads to prediction failure

Sequence independent

Chiti-Dobson

• system is sequence independent
• uses a real-value vector in order to do aggregation prediction
  • < hydrophobicity, β-sheet propensity, charge >
• system takes into consideration composition (amino acid abundance) but not sequence
• create equations modelled on known aggregating proteins
• added more data given 27 mutations -- unstructured proteins for which ThT data available
  • ThT binding -- a well characterized amyloid assay -- tends to bind aggregated proteins
• get calculated rate of ThT fluorescence for mutant and wildtype
• compared calculation (given model) with observation to characterize goodness of model
• first -- observed aggregation value

Agar Method

• next -- measure values against disease duration
• recalibrated Chit-Dobson method --
• used all published mutant-aggregation effect data
• overarching goal is to understand whether current work correlates the ALS/SOD data previously studied
• three sets of data retrieved
  • aggregation propensity
  • instability ( == 1 - normalized ΔΔG or ΔT_m)
  • Sum of aggregation propensity and instability
• remember: we expect instability to correlate with fatality
  • i.e. unfolded assists in aggregation

Zyggregator

• sequence-based prediction algorithm
• uses prediction terms of amino acid secondary structure
• -- considers a sliding window of seven
• aggregation in vitro
• additional terms -- increased weighting toward certain patterns
• -- gatekeeper residues (glycines? polar? -- start or end single aa) hinders aggregation
• three ways to use the prediction algorithm
  1. Ztox : unfolded state → oligomeric state
  2. Zagg : unfolded state → amyloid state
  3. Zagg + CamP : amyloid state → native state
• 1, 2: use same method, parameters fitted to database -- chains may aggregate or into amyloid or protofibrillar
• ...

TANGO

• sliding window of 5aa
• if a window scores over p() = 5%, this is considered aggregation prone
• See also: WALTZ
• See also: Profile 3D
• designed with conformation sites in mind β-turn;, α-helix, β-aggregate.

PASTA

• runs pairs of 7aa sliding window
• parallel or anti-parallel -- calculate energy
• lowest energy matched regions are predicted for high aggregation propensity
• why 7aa? 6aa is the minimum amount required for β aggregates -- 7 gives us a centroid

FoldAmyloid

• a bit different than the other software
• amyloid prediction based on hydrogen formation when residues are packed
• used structural database as an input
• considered contacts
• description of hydrogen bonds with respect to amino acid -- either a donor or an acceptor
• -- criticism: unable to control sliding window size / sliding window size not known
• threshold used to determine which regions are amyloid prone with respect to each sliding window
• model has three pieces
  1. contacts
  2. h-bond donors
  3. h-bond acceptors
• structure implicitly considered -- core aa tend to be hydrophobic for instance

Including structural information

WALTZ

• uses AmylHex database -- experimentally verified hexamer database
  • 67 positive amyloid forming
  • 91 negative for amyloid formation
• uses this database as a starting point
• construct 49 new amyloidogenic, 71 non-amyloidogenic hexapeptides -- experimentally verified
• position specific pseudo-energy matrix
• S_total = α_profileS_profile+α_physpropS_physprop+α_structS_struct
• FoldX used for energy-optimization
• background
  • glycine good for aggregation - flexible
  • aspartate, glutamate bad for aggregation - burying a negative charge
• WALTZ group also wrote TANGO
  • WALTZ is amyloid specific -- not any other aggregation
  • amyloid is position dependent

ZipperDB

• Eisenberg's group
• 3D profiling
• sequence is threaded onto backbone of known amyloid forming hexapeptide -- NNQQNY
• predictions verified in vitro
• RosettaDesign potential energy function used
• Profile 3D -- next step -- starts off with a bunch of hexapeptides instead of just one (NNQQNY)

Predicting SOD Aggregation

• ALS is not considered an amyloid disease
  • amyloid not found in vivo in ALS patients
• SOD documented forming amyloid-like aggregates in vitro