(lower perplexity = the model is more confident the peptide is a good binder)
Part 2: AlphaFold3 structure prediction
For each result, I looked at the below:
The ipTM score (higher is better, closer to 1.0 means confident binding prediction)
Binder
Pseudo Perplexity
ipTM score
WLSYPVVLEWGE
16.4334209
0.21
WHYYVVAVRWKE
31.112519
0.35
WHSYAAAAALWE
12.9540785
0.27
WLYYAVGLAWKX
14.2779037
Could not be generated because it had a letter X
FLYRWLPSRRGG
XXXX
0.3
The ipTM score (interface predicted TM-score) measures how confidently AlphaFold3 predicts the two chains interact. It ranges from 0 to 1.
Below 0.4 = poor
0.4–0.6 = moderate
Above 0.6 = good
Where the peptide appears to dock (look at the 3D viewer — is it near position 4 at the N-terminus? At the dimer interface? Surface-exposed?)
I have a question on the fact that my second binder has a high Pseudo Perplexity value of 31 therefore the assumption is that the binding confidence level is very low. However through alphafold it has the best ipTM score suggesting that it had the best confident binding prediction
Part 3: PeptiVerse therapeutic properties
Binder
Solubility
Hemolysis
Binding Affinity
Length
Molecular Weight
Net Charge (pH 7)
Isoelectric Point
Hydrophobicity (GRAVY)
WLSYPVVLEWGE
Soluble - 1.00
Non-hemolytic 0.112
Weak binding - 5.952
12
1477.7
-2.23
4.24
0.26
WHYYVVAVRWKE
Soluble - 1.00
Non-hemolytic 0.069
Weak binding - 6.345
12
1635.9
0.85
8.5
-0.42
WHSYAAAAALWE
Soluble - 1.00
Non-hemolytic 0.035
Weak binding - 5.863
12
1375.5
-1.15
5.47
0.18
WLYYAVGLAWKX
Soluble - 1.00
Non-hemolytic 0.115
Medium binding - 7.101
12
1351.8
0.76
8.5
0.56
the best one balances strong binding + not hemolytic + good solubility.
I initially started with trying to use the google colab shared and input what I thought was right. I continously kept on getting an error that is IndexError.
I used claude to help me work out what i can do, and this is what it provided me with the instructions below.
After that I did the following and this is a summary of my concluison from the work.
Steps taken
I then went to Gemini and copy pasted the above and let it run.
After completing the run the output it gave was as below
moPPIt uses discrete flow matching to steer generation toward specific binding sites AND simultaneously optimize multiple properties.
moPPIt peptides are expected to have higher binding affinity and better drug-like properties (solubility, safety) than PepMLM outputs
SOD1 Structure Guide: Targeting Residues for moPPIt
What Is SOD1?
Superoxide Dismutase 1 (SOD1)
Role: Antioxidant enzyme in cells; protects against free radical damage
Structure: Homodimer (~32 kDa monomer × 2) with zinc and copper cofactors
Disease Link: >180 mutations cause Familial ALS (fALS); A4V is one of the most severe
A4V Mutation Impact:
Position 4: Alanine → Valine
Makes protein unstable, prone to misfolding and aggregation
Leads to neuronal toxicity → motor neuron death → ALS
Position: 1 2 3 4 5 6 7 8 9 10
Sequence: M E T A→V K S Q V V Q
Index: 0 1 2 3 4 5 6 7 8 9
↑↑↑↑↑↑↑↑↑↑↑↑↑
TARGET INDEX: 3
Why target here?
✓ Direct site of pathogenic mutation
✓ Likely destabilized region
✓ Peptide here could stabilize or mark for degradation
✓ Highest specificity to A4V disease
Recommended for: Stabilization or selective targeting of mutant
2. Metal Cofactor Binding Sites (Stabilization)
Zinc Binding (Zn²⁺):
Position: 45 46 47 48 49
Sequence: R P D E D
Index: 44 45 46 47 48
↑ ↑↑↑↑↑↑↑↑↑↑↑↑↑
| Zn²⁺ binding motif
Loop
Why target here?
✓ Zinc stabilizes SOD1 structure
✓ A4V mutants have weaker Zn²⁺ binding
✓ Peptide here could enhance Zn²⁺ coordination
✓ Stabilizes fold → prevents aggregation
Recommended indices: [44, 45, 46, 47, 48]
Or 1-indexed positions: 45-49
Copper Binding (Cu²⁺):
Position: 70 71 72 73 74 75 76 77 78 79 80
Sequence: H S V Y V D Q W D W E
Index: 69 70 71 72 73 74 75 76 77 78 79
↑ ↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑
| Cu²⁺ binding cluster
Loop
Key residues:
- H71 (Histidine 71) - coordinates Cu²⁺
- D74 (Aspartate 74) - coordi
nates Cu²⁺
- W76 (Tryptophan 76) - structural support
Why target here?
✓ Critical for catalytic activity
✓ A4V mutants lose Cu²⁺ stability faster
✓ Peptide here could improve metal retention
✓ Even small improvements help
Recommended indices: [70, 71, 73, 74, 76, 78, 79]
Or 1-indexed positions: 71, 72, 74, 75, 77, 79, 80
3. Dimer Interface (Aggregation Prevention)
MONOMER A: MONOMER B:
┌──────────────────┐ ┌──────────────────┐
│ │ │ │
│ Position 50-60 │◄─────────►│ Position 50-60 │
│ (contacts B) │ Interface │ (contacts A) │
│ │ │ │
│ Position 85-100 │◄─────────►│ Position 85-100 │
│ (contacts B) │ Interface │ (contacts A) │
│ │ │ │
└──────────────────┘ └──────────────────┘
Dimer Interface A (Region 1):
Position: 50 51 52 53 54 55 56 57 58 59 60
Sequence: L G Q H D F S A G E G
Index: 49 50 51 52 53 54 55 56 57 58 59
↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑
Dimer Interface B (Region 2):
Position: 85 86 87 88 89 90 91 92 93 94 100
Sequence: G I E Q L P D G Q K ...
Index: 84 85 86 87 88 89 90 91 92 93 99
↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑
Why target here?
✓ A4V mutants hyperaggregate (misfolded dimers)
✓ Peptide here could disrupt "bad" dimers
✓ Or stabilize native dimers
✓ Critical for ALS pathogenesis
Recommended indices: [49, 51, 53, 55, 57, 84, 86, 88, 90]
Or 1-indexed positions: 50, 52, 54, 56, 58, 85, 87, 89, 91
4. General Surface Patches (High Accessibility)
SOD1 accessible surface:
- N-terminal region (1-20): Accessible, flexible
- Loop regions (40-50, 60-70): Flexible, exposed
- C-terminal region (140-153): Accessible, flexible
- Anywhere NOT in active site or dimer interface
Benefits:
✓ Easier for peptide to access
✓ Less steric clashes
✓ Multiple contact points
Recommended indices: [0-20] or [100-153] or scattered accessible residues
Goal: Keep SOD1 folded (prevent A4V misfolding)
Strategy: Reinforce metal binding sites
Biology positions: 45 46 47 48 49 70 71 72 73 74 75 76
Code indices: 44 45 46 47 48 69 70 71 72 73 74 75
For Colab input:
binding_residue_indices = [44, 45, 46, 47, 48, 69, 70, 71, 72, 73, 74, 75]
Explanation: "I'm targeting the Zn²⁺ and Cu²⁺ coordination sites. A4V
mutants are unstable, so a peptide that reinforces these metal-binding
regions could restore structural integrity."
Example 2: Anti-Aggregation Strategy
Goal: Prevent dimerization (block aggregation)
Strategy: Disrupt dimer interface
Biology positions: 50 52 54 56 58 85 87 89 91
Code indices: 49 51 53 55 57 84 86 88 90
For Colab input:
binding_residue_indices = [49, 51, 53, 55, 57, 84, 86, 88, 90]
Explanation: "I'm targeting the dimer interface. A4V causes pathogenic
aggregates, so a peptide at the dimer interface could selectively bind
and disaggregate or prevent formation of toxic SOD1 oligomers."
Example 3: Multi-Site Strategy (Recommended)
Goal: Multi-pronged approach
Strategy: Combine mutation site + metal binding + accessible surface
Targets:
- A4V site: position 4 → index 3
- Zn binding: positions 45-49 → indices 44-48
- Cu binding: positions 71, 74, 76 → indices 70, 73, 75
- Surface accessibility: position 100 → index 99
For Colab input:
binding_residue_indices = [3, 44, 45, 46, 47, 48, 70, 73, 75, 99]
Explanation: "I'm targeting multiple sites: the A4V mutation site (direct
targeting), the Zn²⁺ and Cu²⁺ binding regions (structural stabilization),
and a surface-accessible region (for cell recognition). This multi-target
approach should generate peptides with both high affinity and therapeutic
breadth."
INDEX CONVERSION QUICK TABLE
If you see position X in publications/UniProt:
Biology Position
Code Index
Region
4
3
A4V mutation
45-49
44-48
Zn²⁺ site
50-60
49-59
Dimer interface 1
71-80
70-79
Cu²⁺ site
85-100
84-99
Dimer interface 2
1-20
0-19
N-terminus
140-153
139-152
C-terminus
Formula: Code Index = Biology Position - 1
SAMPLE OUTPUT FROM moPPIt
Once you run generation, you’ll see output like this:
