HemoPI-MOD2 predicts the hemolytic potency (HC50 in µM) of natural and chemically modified peptides from MAP sequence, SMILES strings, or 3D PDB structures, and includes a single-residue analog-design module. This page explains how to read prediction and design results, how to write modified sequences in MAP notation, and how to build a command for the standalone CLI.
HemoPI-mod2 runs regression models that predict pHC50 and report HC50 in µM (HC50 = 10−pHC50), alongside a Hemolytic or Non-Hemolytic label using a fixed 100 µM cut-off. Match the numbered markers in the figure against the list underneath.
Which prediction model produced this run (Sequence, Chemical, or Structure) and the submitted input format (MAP, SMILES & PDB).
Downloads the full prediction table as a CSV file.
Orders rows by predicted HC50. Click again to reverse the direction.
Type to show only rows whose ID matches. Useful for large batches; sorting and filtering work together.
The predicted HC50 in µM (computed from pHC50 as 10−pHC50).
Hemolytic or Non-Hemolytic. HC50 ≤ 100 µM is labelled Hemolytic; above 100 µM, Non-Hemolytic.
The design module performs a single-site saturation scan on one input peptide, substituting each residue position with every other canonical amino acid to produce 19 × N variants alongside the wild-type (WT) reference. Existing MAP modifications are preserved in every variant; only the single residue at the scanned position changes.
The on-screen design table displays four columns:
Variant,
Sequence,
Prediction (HC50), and
Prediction. Results are paginated at
10 rows per page with First, Prev, Next, and
Last controls. Clicking the Variant or
PREDICTION (HC50) column header cycles sorting
across the entire scan (ascending, descending, then back to original
residue order) while keeping the wild-type row pinned at the top.
The downloadable CSV file contains the complete scan
with all nine columns: Variant,
Position, Wild_Residue,
Mutant_Residue, Sequence,
pHC50, HC50_uM,
Fold_Change_vs_WT, and
Direction.
MAP (Modification & Annotation in Proteins) notation writes a chemically
modified peptide as its one-letter amino-acid sequence with modification
tokens in curly braces. Standard residues are written normally; a
token in { } either modifies the residue just
before it, or sets a property of the whole peptide.
For example, ACA{d}DEFKHIK{ct:amid} reads as the
sequence ACADEFKHIK where the third residue (the A before
{d}) is in D-configuration, and the C-terminus
carries an amide cap. Unmodified peptides need no tokens at all — a plain
string like GIGKFLHSAKKFGKAFVGEIMNS is valid MAP.
| Token | Applies to | Meaning |
|---|---|---|
{nt:acet} | N-terminus | Caps or modifies the N-terminal end for acetylation. |
{ct:amid} | C-terminus | Caps or modifies the C-terminal end for amidation. |
{d} | Preceding residue | Marks the residue immediately before it as the D-enantiomer. |
{nnr:orn} | Own position | Inserts a non-natural residue at that position in the chain, named by its code (orn = ornithine). |
{cyc:n-c} | Whole peptide | Head-to-tail (backbone) cyclization joining the N- and C-termini. |
ct:Amid, cyc:N-C,
d, nnr:orn,
nt:Acet), and 3 stereochemistry summaries
(total_D_count,
total_L_count, D_ratio)
from MAP sequences.
Chemical model (XGBoost, 100 2D Mordred descriptors): accepts MAP
sequences (converted to SMILES internally) or SMILES strings directly.
Structure model (ExtraTrees, 100 2D Mordred descriptors): computes
descriptors from 3D PDB coordinates (e.g. from PEPstrMOD2).
KWKLFKKIGAVLKVL{cyc:n-c}
A 14-residue peptide, cyclized head-to-tail (N-terminus bonded to C-terminus).
ACA{d}DEFKHIK{ct:amid}
Sequence ACADEFKHIK with a D-alanine at position 3 and a C-terminal amide.
GIGKFLHSAKKFGKAFVGEIMNS
A plain natural peptide (magainin-like) — no tokens, valid MAP as-is.
{nt:acet}FL{nnr:ORN}KVLK
Contains a non-natural residue (ornithine) and an N-terminal acetyl cap.
FASTA / sequence input (Sequence or Chemical model): paste one or
more records in FASTA format, each with a header line beginning with
> followed by the MAP sequence:
>Pep2
KWKLFKKIGAVLKVL{cyc:n-c}
>Pep3
GIGKFLHSAKKFGKAFVGEIMNS
Headers you provide are kept as-is, while only
> with no text produces an empty identifier.
If you omit FASTA headers altogether (one sequence per line, no
>), records are auto-named
seq_1, seq_2, and so on.
SMILES input (Chemical model): enter one SMILES string per line,
or use FASTA format with a > header line to
name each molecule.
PDB input (Structure model): paste the full PDB coordinate text, one structure per submission; all-atom structures including hydrogens are accepted.
Prefer to run HemoPI-mod2 offline? Choose your options
below and the exact HemoPI_MOD2.py command appears in
the console. Copy it and run it in your terminal.
Path to .fasta, .fa, or .txt input (for model 3, this may be a .pdb file or a folder of them).
Output CSV path (default: output.csv).
Run from the standalone directory containing
HemoPI_MOD2.py. Optional flags:
-d (1 shows only hemolytic
peptides on screen, 2 shows all) and
-wd (working directory).