pip install ablang

import ablang
# 初始化重鏈模型
heavy_model = ablang.pretrained("heavy")  # 重鏈模型
# 初始化輕鏈模型
light_model = ablang.pretrained("light")  # 輕鏈模型

# 生成序列嵌入向量
seq = "EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYGMHWVRQAPGKGLEWVSAISGSGGSTYYADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAKDRGGYEFDFWGTGTLVTVSS"
embedding = heavy_model(seq, mode="embed")

# 預測突變效應
wt_seq = "QVQLQESGPGLVAPSQSLSITCTVSGFSLTNYGVHWVRQPPGKGLEWLGVIWGSETTYYNSALKSRLTIIKDNSKSQVFLKMNSLQTDDTAIYYCAKVKFYDPTAPNDYWGQGTLVTVSS"
mutations = ["S30A", "H35Y", "D100E"]  # 格式: 野生型氨基酸+位置+突變氨基酸
# 預測突變對穩(wěn)定性的影響（ΔΔG）
predictions = heavy_model(wt_seq, mode="predict", mutations=mutations)

import ablang
import numpy as np
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
# 初始化模型
heavy = ablang.pretrained("heavy")
# 抗體序列列表
antibody_sequences = [
    "EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYGMHWVRQAPGKGLEWVSAISGSGGSTYYADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAKDRGGYEFDFWGTGTLVTVSS",
    "QVQLQESGPGLVAPSQSLSITCTVSGFSLTNYGVHWVRQPPGKGLEWLGVIWGSETTYYNSALKSRLTIIKDNSKSQVFLKMNSLQTDDTAIYYCAKVKFYDPTAPNDYWGQGTLVTVSS",
    # 更多序列...
]
# 生成嵌入向量
embeddings = []
for seq in antibody_sequences:
    emb = heavy(seq, mode="embed")
    embeddings.append(emb.detach().numpy())  # 轉換為numpy數(shù)組
# K-means聚類
kmeans = KMeans(n_clusters=3, random_state=42)
clusters = kmeans.fit_predict(embeddings)
# 可視化聚類結果（使用PCA降維）
from sklearn.decomposition import PCA
pca = PCA(n_components=2)
pca_result = pca.fit_transform(embeddings)
plt.scatter(pca_result[:, 0], pca_result[:, 1], c=clusters)
plt.title("Antibody Sequence Clustering")
plt.xlabel("PCA Component 1")
plt.ylabel("PCA Component 2")
plt.show()

import ablang
# 初始化重鏈模型
heavy = ablang.pretrained("heavy")
# 野生型抗體序列
wt_sequence = "EVQLQESGPGLVAPSQSLSITCTVSGFSLTNYGVHWVRQPPGKGLEWLGVIWGSETTYYNSALKSRLTIIKDNSKSQVFLKMNSLQTDDTAIYYCAKVKFYDPTAPNDYWGQGTLVTVSS"
# 設計潛在的穩(wěn)定性增強突變
potential_mutations = [
    "S30A", "H35Y", "T50I", "N58D", 
    "K62R", "L89V", "A92P", "V100L"
]
# 預測突變效應
results = heavy(wt_sequence, mode="predict", mutations=potential_mutations)
# 輸出結果并排序（尋找最有利的突變）
sorted_results = sorted(zip(potential_mutations, results), 
                       key=lambda x: x[1], reverse=True)
print("突變穩(wěn)定性預測結果（從高到低）:")
for mut, score in sorted_results:
    print(f"{mut}: {score:.4f}")

import ablang
import random
# 初始化輕鏈模型
light = ablang.pretrained("light")
# 起始抗體序列
starting_sequence = "DIQMTQSPSSLSASVGDRVTITCRASQDVSTAVAWYQQKPGKAPKLLIYSASFLYSGVPSRFSGSRSGTDFTLTISSLQPEDFATYYCQQHYTTPPTFGQGTKVEIK"
# 定義優(yōu)化位置
optimization_positions = [30, 31, 50, 53, 95]  # 要優(yōu)化的氨基酸位置
# 隨機突變并評估
best_sequence = starting_sequence
best_score = light(best_sequence, mode="predict")  # 基礎分數(shù)
# 簡單的隨機搜索優(yōu)化
for _ in range(100):  # 嘗試100種突變組合
    new_seq = list(starting_sequence)
    # 在指定位置引入隨機突變
    for pos in optimization_positions:
        # 隨機選擇20種天然氨基酸之一
        new_aa = random.choice("ACDEFGHIKLMNPQRSTVWY")
        new_seq[pos] = new_aa
    new_seq = "".join(new_seq)
    # 評估新序列
    new_score = light(new_seq, mode="predict")
    # 如果更好則更新
    if new_score > best_score:
        best_score = new_score
        best_sequence = new_seq
print(f"優(yōu)化后的序列: {best_sequence}")
print(f"優(yōu)化后的分數(shù): {best_score:.4f}")

import ablang
import numpy as np
from scipy.spatial.distance import cosine
# 初始化模型
heavy = ablang.pretrained("heavy")
# 比較兩個抗體序列的相似性
seq1 = "EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYGMHWVRQAPGKGLEWVSAISGSGGSTYYADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAKDRGGYEFDFWGTGTLVTVSS"
seq2 = "QVQLQESGPGLVAPSQSLSITCTVSGFSLTNYGVHWVRQPPGKGLEWLGVIWGSETTYYNSALKSRLTIIKDNSKSQVFLKMNSLQTDDTAIYYCAKVKFYDPTAPNDYWGQGTLVTVSS"
# 獲取嵌入向量
emb1 = heavy(seq1, mode="embed").detach().numpy()
emb2 = heavy(seq2, mode="embed").detach().numpy()
# 計算余弦相似度
similarity = 1 - cosine(emb1, emb2)
print(f"序列相似度: {similarity:.4f}")  # 值越接近1表示越相似

import ablang
import pandas as pd
# 初始化模型
heavy = ablang.pretrained("heavy")
# 從CSV文件加載抗體庫（假設包含"sequence"列）
antibody_library = pd.read_csv("antibody_library.csv")
# 批量處理并評分
scores = []
for seq in antibody_library["sequence"]:
    try:
        # 過濾無效序列
        if len(seq) < 100 or len(seq) > 150:
            scores.append(None)
            continue
        # 預測序列質量分數(shù)
        score = heavy(seq, mode="predict")
        scores.append(score)
    except:
        scores.append(None)
# 添加評分到DataFrame
antibody_library["predicted_score"] = scores
# 篩選前100個高分序列
top_candidates = antibody_library.dropna().sort_values(
    by="predicted_score", ascending=False).head(100)
# 保存結果
top_candidates.to_csv("top_antibody_candidates.csv", index=False)
print(f"篩選完成，保存了{len(top_candidates)}個候選序列")

import ablang
import matplotlib.pyplot as plt
# 初始化重鏈模型
heavy = ablang.pretrained("heavy")
# 不同種系來源的抗體序列
germline_sequences = {
    "IGHV1-69": "QVQLVQSGAEVKKPGASVKVSCKASGYTFTNYYMYWVRQAPGQGLEWMGGINPSNGGTNFNEKFKNRVTLTTDSSTTTAYMELRSLRSDDTAVYYCAR",
    "IGHV3-23": "EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYGMHWVRQAPGKGLEWVSAISGSGGSTYYADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAK",
    "IGHV4-34": "QVQLQQSGAELVKPGASVKISCKASGYTFTSYWMHWVRQAPGQGLEWMGWINPNSGGTNYAQKFQGRVTMTTDTSTSTAYMELRSLRSDDTAVYYCAR"
}
# 分析種系序列特征
embeddings = {}
for name, seq in germline_sequences.items():
    embeddings[name] = heavy(seq, mode="embed").detach().numpy()
# 計算種系間的距離
distances = {}
germlines = list(germline_sequences.keys())
for i in range(len(germlines)):
    for j in range(i+1, len(germlines)):
        dist = cosine(embeddings[germlines[i]], embeddings[germlines[j]])
        distances[f"{germlines[i]} vs {germlines[j]}"] = dist
# 可視化距離
plt.bar(distances.keys(), distances.values())
plt.title("Germline Sequence Distances")
plt.ylabel("Cosine Distance")
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()

model = ablang.load_model("/path/to/downloaded/model")

model = ablang.pretrained("heavy", device="cpu")