Bioinformatique

from Bio import SeqIO
from Bio.Seq import Seq
from Bio.SeqUtils import gc_fraction

# Lecture d'un fichier FASTA
for record in SeqIO.parse("genome.fasta", "fasta"):
    seq = record.seq
    print(f"ID: {record.id} — Longueur: {len(seq)} pb")
    print(f"GC content : {gc_fraction(seq):.2%}")

# Opérations de base
dna = Seq("ATGGCCATTGTAATGGGCCGCTGAAAGGGTGCCCGATAG")
print(dna.complement())        # Brin complémentaire
print(dna.reverse_complement()) # Brin antisens 5'→3'

# Traduction ADN → ARNm → protéine
mrna = dna.transcribe()
prot = mrna.translate(to_stop=True)
print(f"Protéine : {prot}")

# Comptage des k-mers (k=4)
k, kmers = 4, {}
for i in range(len(dna) - k + 1):
    km = str(dna[i:i+k])
    kmers[km] = kmers.get(km, 0) + 1
print(sorted(kmers.items(), key=lambda x: -x[1])[:5])

from Bio.PDB import PDBParser, DSSP
import numpy as np

# Chargement d'une structure PDB
parser    = PDBParser(QUIET=True)
structure = parser.get_structure("hemo", "1hho.pdb")
model     = structure[0]

# Structure secondaire via DSSP
dssp  = DSSP(model, "1hho.pdb")
helix = sum(1 for k in dssp if dssp[k][2] == 'H')
sheet = sum(1 for k in dssp if dssp[k][2] == 'E')
print(f"Hélices α : {helix} résidus — Feuillets β : {sheet} résidus")

# Coordonnées Cα + rayon de giration
ca = np.array([
    res['CA'].get_vector().get_array()
    for chain in model
    for res   in chain
    if 'CA' in res
])
center = ca.mean(axis=0)
rg     = np.sqrt(((ca - center)**2).sum(axis=1).mean())
print(f"Rayon de giration Rg = {rg:.2f} Å")

import torch
import esm  # pip install fair-esm

# ESM-2 — 150M paramètres
model, alphabet = esm.pretrained.esm2_t30_150M_UR50D()
batch_converter  = alphabet.get_batch_converter()
model.eval()

sequences = [
    ("signal_peptide", "MKTIIALSYIFCLVFA"),
    ("hemoglobin_alpha", "MVLSPADKTNVKAAW"),
]

_, _, tokens = batch_converter(sequences)

with torch.no_grad():
    out = model(tokens, repr_layers=[30], return_contacts=True)

# Embeddings [B, L, 640] — couche 30
reprs = out["representations"][30]
for i, (label, seq) in enumerate(sequences):
    mean_emb = reprs[i, 1:len(seq)+1].mean(dim=0)
    print(f"{label}: shape={mean_emb.shape}")

# Matrice de contacts prédite [B, L, L]
contacts = out["contacts"]
print(f"Contact map: {contacts.shape}")