
中国科学院微生物研究所,中国微生物学会
文章信息
- 汪庆, 张瑞芬, 王亚楠, 朱宝利, 曾斌. 2022
- WANG Qing, ZHANG Ruifen, WANG Yanan, ZHU Baoli, ZENG Bin.
- 抗菌肽结构改造与人工智能研发策略
- Antimicrobial peptides: structure modification and development with artificial intelligence
- 微生物学报, 62(11): 4353-4366
- Acta Microbiologica Sinica, 62(11): 4353-4366
-
文章历史
- 收稿日期:2022-03-17
- 修回日期:2022-07-29
- 网络出版日期:2022-08-11
2. 江西科技师范大学化学化工学院, 江西 南昌 330013;
3. 中国科学院病原微生物与免疫学重点实验室, 北京 100101;
4. 深圳技术大学药学院, 广东 深圳 518118
2. College of Chemistry and Chemical Engineering, Jiangxi Science and Technology Normal University, Nanchang 330013, Jiangxi, China;
3. CAS Key Laboratory of Pathogenic Microbiology and Immunology, Institute of Microbiology, Chinese Academy of Sciences, Beijing 100101, China;
4. College of Pharmacy, Shenzhen Technology University, Shenzhen 518118, Guangdong, China
抗菌肽(antimicrobial peptides,AMPs)是生物体在长期进化过程中为适应环境而产生的免疫活性分子,具有抗菌、抗病毒、抗真菌和抗寄生虫等功能,在机体的天然免疫防御系统中发挥着重要的作用,也被称为“宿主防御肽”(host defence peptides,HDPs)。抗菌肽通常由12–50个氨基酸残基组成,其中绝大部分是阳离子短肽,电荷携带量在+2–+9之间,带有+2–+4电荷的抗菌肽最为丰富[1−2]。抗菌肽表面携带大量正电荷原因在于,抗菌肽与细菌细胞膜的初始相互作用主要是通过抗菌肽表面的阳离子残基与细菌细胞膜上阴离子的静电作用来实现。另外,也含有少量阴离子抗菌肽,但阴离子抗菌肽抗菌活性远低于阳离子抗菌肽,且阴离子抗菌肽在先天性免疫中的作用还尚未完全了解,因此对于抗菌肽的研究主要集中于阳离子抗菌肽[3]。结构上,抗菌肽具有α-螺旋、β-折叠、延伸/随机卷曲等3种构型,这些结构具有两亲性或具有转变成两亲性的能力,使抗菌肽能溶于水和富含脂质的环境中[4],从而可以粘附到细菌细胞膜表面并发挥膜溶解作用。抗菌肽的分子量也较小,通常小于10 kDa,小的分子量有助于抗菌肽在宿主细胞中快速扩散和分泌,激发对病原微生物的即时防御[3]。
抗菌肽最早由青霉素的发现者亚历山大·弗莱明于1922年在鼻腔分泌物中发现,而当时他把这种具有抗菌活性的蛋白称为“溶菌酶”[5],尽管溶菌酶的抗菌依赖于细菌细胞内酶的作用,而抗菌肽利用非酶作用机制,但该“溶菌酶”是第一个在动物细胞中发现的具有抗菌活性的多肽。此后弗莱明也在其他动植物组织、体液中发现这种抗菌蛋白的存在,表明该抗菌蛋白作为机体免疫系统的一部分具有广泛的功能[1]。1939年René Dubos从土壤细菌短芽孢杆菌Bacillus brevis的上清液中分离出第1个抗菌肽gramicidin,这是第1个商业化生产的抗菌肽[6]。但由于同时期抗生素的广泛使用,导致抗菌肽的有关研究并没有受到普遍关注。直至20世纪60年代,抗生素的广泛使用导致耐药性病原体逐渐增加,抗菌肽凭借广谱抗菌能力和不易产生耐药性等特点,开始受到人们的广泛关注。
1 抗菌肽的抗菌机制抗菌肽的抗菌机制可以概括为两大类:细胞膜靶向作用和非膜靶向作用,其中细胞膜靶向作用是抗菌肽最常见的作用机制,非膜靶向作用是指抗菌肽能进入细菌细胞内,保持细胞膜完整性的同时在胞内积累,对细菌关键代谢过程产生影响,从而达到抗菌效果。
1.1 细胞膜靶向作用机制细胞膜靶向作用实现细菌细胞杀伤可分为3个步骤。抗菌肽与靶细胞相互吸引、抗菌肽与细胞膜中的脂质相互作用和抗菌肽介导细胞膜破坏。
(1) 抗菌肽与靶细胞相互吸引。带有正电荷的抗菌肽会被革兰氏阴性菌外膜上的脂多糖(lipopolysaccharides,LPS)和革兰氏阳性菌细胞壁上的脂磷壁酸(lipoteichoic acids,LTA)所吸引,通过静电相互作用结合到细菌表面[7]。由于抗菌肽具有两亲性,其带正电荷的极性面驱动对细胞膜上带负电部位静电吸引,然后抗菌肽的非极性面通过疏水力和范德华力相互作用,使抗菌肽插入细菌细胞的脂质双分子层中[8]。
(2) 抗菌肽与细胞质膜中脂质相互作用。对于革兰氏阳性菌,抗菌肽需要穿过荚膜多糖(capsular polysaccharides,CPS)、磷壁酸(teichoic acids,TA)以及LTA与细胞质膜结合并相互作用[9],而革兰氏阴性菌细胞膜表面含有孔蛋白,抗菌肽穿过孔蛋白与细胞质膜结合。一旦通过细菌细胞壁或外膜,抗菌肽会与细菌细胞质膜中带负电荷的脂质相互作用,如磷脂酰甘油(phosphatidylglycerols,PG)、磷脂酰丝氨酸(phosphatidylserines,PS)和心磷脂(cardiolipins,CL)[10−11],取代Ca2+、Mg2+等稳定磷脂的二价阳离子,从而导致膜结构的破坏[12]。由于哺乳动物的细胞膜中主要存在中性的磷脂酰胆碱(phosphatidylcholines,PC)、磷脂酰乙醇胺(phosphatidylethanolamines,PE)、鞘磷脂(sphingomyelins,SM)等,PG、PS、CL很少或几乎不存在[13−14]。因此,抗菌肽可以特异性地杀死病原体[15]。另外,也有研究表明,电荷不是介导细胞膜脂质与抗菌肽相互作用的唯一因素,细胞膜的“膜曲率”也发挥着重要作用。膜曲率指磷脂双分子层上脂质的排列角度,生物膜上的脂质以一定的角度排列,这是生物膜得以弯曲的基础[16]。由于脂质的形状取决于其头部和疏水尾的相对大小,PG与PS分子形状分别为圆柱形和圆锥形,显示出不同的膜曲率特性。抗菌肽如magainins能更有效地诱导由PG组成的脂质体的渗漏,而对哺乳动物细胞膜无影响[17]。
(3) 抗菌肽介导细胞膜破坏。由于抗菌肽与细菌细胞膜之间的静电相互作用,使抗菌肽在细菌细胞膜上累积,达到一定浓度后,抗菌肽改变自身结构破坏细菌细胞膜而发挥抗菌作用,导致细菌细胞膜去极化、膜电位下降、膜通透性改变、细胞内大分子物质泄漏,最终导致细菌死亡[18]。目前,已提出了几种模型来解释抗菌肽介导的细菌细胞膜结构破坏机制,分别是桶壁模型、地毯模型以及环孔模型。3种模型区别与示意图如表 1、图 1与图 2所示。
Models | Characteristics of interaction with bacterial cell membrane | Peptides | References |
Barrel-stave | Antimicrobial peptides are inserted into the membrane by orienting their hydrophobic regions in the core of lipid bilayer to form barrel shaped transmembrane pores. The model is characterized by the vertical aggregation of helices into the lipid bilayer | Alamethicin; pardaxin; bacSp222; Class II bacteriocins | [19–23] |
Carpet | The peptide is electrostatically attracted to the anionic phospholipid head group on the membrane surface in a carpet like manner, and destroys the membrane structure in the form of detergent. At the critical threshold concentration, the peptide forms a circular transient pore on the membrane, allowing additional peptides to enter the membrane | Cecropins; dermaseptin; aurein 1.2; LL-37 | [9, 24–26] |
Toroidal-pore | The peptides inserted into the membrane cause a continuous bending of the lipid monolayer from top to bottom, the loose interaction between the polar group of the peptide and the phospholipid head on the bacterial membrane leads to the formation of pores randomly arranged by hydrophilic groups. This structure is similar to the barrel-stave model, but the pores formed are instantaneous and more unstable than the barrel-stave structure | Maganinis; lacticin Q; arenicin; aurein 2.2; melittin; helical PGLa; buforin II | [26–29] |
![]() |
图 1 抗菌肽细胞膜靶向作用机制示意图 Figure 1 Schematic diagram of cell membrane targeting mechanism of antimicrobial peptides. For Gram-positive bacteria, antimicrobial peptides first attract each other with LTA on the cell wall, and then combine with the cell plasma membrane through LTA to replace the cations that stabilize phospholipids, resulting in the destruction of the cell membrane, as shown in the left figure. For Gram-negative bacteria, antimicrobial peptides are attracted to the cell surface by LPS on the outer membrane, pass through the porin on the outer membrane and bind to the plasma membrane to replace the cations that stabilize the phospholipids, resulting in the destruction of the membrane structure, as shown in the right figure. |
![]() |
图 2 抗菌肽介导细胞膜结构破坏的机制示意图 Figure 2 Schematic diagram of the mechanism of antimicrobial peptides mediated destruction of cell membrane structure. When antimicrobial peptides bind to the plasma membrane, they destroy the cell membrane structure through the three modes of barrel-stave, carpet and toroidal-pore, and finally lead to the lysis of bacterial cells. The red part of the surface of the antimicrobial peptides represents the non-polar surface of the antimicrobial peptides, and the blue part represents the polar surface. |
1.2 非膜靶向作用机制
非膜靶向作用的抗菌肽根据其作用靶点可分为两大类:细菌细胞壁靶向的抗菌肽和细菌细胞内靶向的抗菌肽[17]。抗菌肽通常与细胞壁的前体分子相互作用以抑制细胞壁形成,如高度保守的脂质Ⅱ。研究发现,人α-defensins 1 (HNP-1)、β-defensins 3 (HBD-3)可以与脂质Ⅱ选择性结合以阻止细菌细胞壁合成[15]。由革兰氏阳性菌乳酸乳球菌(Lactococcus lactis)产生,广泛用作食品防腐剂的乳酸链球菌素(lantibiotic nisin)也可作用于脂质II,阻止中央细胞壁前体并入正在生长的肽聚糖网络,从而抑制功能性细胞壁形成[30]。
作用于细胞内靶点的抗菌肽一般进入细胞内与细胞内靶点相结合,干扰细胞重要的代谢过程。如青蛙组蛋白衍生的抗菌肽buforin II,小麦胚乳蛋白分离的抗菌肽puroindoline可与大肠埃希菌(Escherichia coli)的DNA和RNA结合,改变它们在1%琼脂糖凝胶中的电泳迁移率,puroindoline B还可以抑制DNA复制、转录以及翻译等过程[31],而分离于日本鲎血细胞膜的抗菌肽tachyplesin可与DNA小沟结合以抑制细菌生长。另外,存在于比目鱼中的pleurocidin和蛙皮分泌物的dermaseptin等α-螺旋肽、富含脯氨酸和精氨酸的PR-39、indolicidin,以及人α-defensins-1 (HNP-1)等抗菌肽均可阻断E. coli的(3H)胸腺嘧啶核苷、(3H)尿苷和(3H)亮氨酸摄取,抑制DNA、RNA和蛋白质合成[9]。然而实际上,在大量已知的天然或人工设计的抗菌肽中,只有少数几类抗菌肽通过非膜靶向作用机制发挥抗菌活性,在任何条件或任何浓度下都不会损伤靶细胞膜的抗菌肽很少,大多数在浓度远高于其最小抑制浓度(minimum inhibitory concentration,MIC)时,都会造成细胞膜破坏[32]。
2 抗菌肽的应用优势与挑战与传统抗生素相比,抗菌肽应用优势主要包括:(1) 选择性毒性:病原体与宿主之间细胞膜组成的差异是抗菌肽靶向特异性的基础,使得抗菌肽可以特异性入侵病原体[10];(2) 作用迅速:抗菌肽杀灭目标细菌所需的时间比目标细菌倍增时间短得多[9];(3) 有效杀灭持留菌:持留菌是细菌群体中可耐受致死剂量抗生素及其他压力环境而存活的细菌亚群, 是未发生遗传性突变、代谢活性低下且不会被抗菌药物杀伤的细菌细胞[33]。由于抗菌肽的作用靶点通常在细胞膜上,不依赖于细胞代谢活性,故有利于杀灭持留菌[34];(4) 广谱性: 除了常见的革兰氏阴性和革兰氏阳性细菌,抗菌肽对病毒、原生动物和真菌等也有效[35]。(5) 不易产生耐药性:抗菌肽作用于高度保守的细胞膜,几乎不能被细菌修饰而产生耐药性;其次,抗菌肽半衰期短,很难在环境中积累而诱导细菌产生耐药性;另外,抗菌肽作用时间短,与抗生素相比,抗菌肽中间治疗浓度范围小于抗生素,中间治疗浓度范围越小,越不易诱导耐药性产生[4]。目前,抗菌肽在食品、农业、医药等领域均有一定的应用并取得了良好的效果。如在食品方面,抗菌肽在食品保鲜领域运用已有二十多年历史,其中研究最多的是乳酸菌(lactic acid bacteria,LAB)家族产生的细菌素(bacteriocins),细菌素可抑制各种食品腐败菌及致病菌,如肠炎沙门菌(Salmonella enteritidis)、单核细胞增生李斯特菌(Listeria monocytogenes)、产气荚膜梭菌(Clostridium perfringens)和蜡样芽孢杆菌(Bacillus cereus)等,并延长食品的货架期。另外,由植物乳杆菌(Lactobacillus plantarum)产生的植物杀菌素(plantaricins)具有广泛的抗菌谱,可抑制食源性病原菌及乳酸杆菌[36]。在农业方面,抗菌肽可作为禽畜饲料添加剂取代或部分取代抗生素以减少抗生素对禽畜的危害。研究表明[37],给断奶的仔猪添加天然抗菌肽cecropins可以增加仔猪的肠道乳酸杆菌的数量,改善仔猪肠道健康。不仅如此,抗菌肽在临床上也有一定的应用,如indolicidin的合成类似物omiganan,在治疗真菌感染方面有很大潜力,可用于治疗酒渣鼻、青少年痤疮、特应性皮炎、生殖器疣和外阴上皮内瘤变[38]。
尽管抗菌肽应用前景光明,但也存在许多问题,主要体现在3个方面:(1) 药代动力学方面:抗菌肽由于具有红细胞溶血性、口服生物利用度低、对胃肠道蛋白酶降解敏感等问题,导致大多数抗菌肽不能采用口服方式给药;而通过静脉注射等全身给药方式,由于血浆中蛋白水解酶能使抗菌肽快速降解,机体肝脏、肾脏的快速清除作用导致抗菌肽半衰期短,故抗菌肽仅限于局部外用给药。(2) 抗菌活性方面:尽管抗菌肽具有广谱抗菌作用,但与传统抗生素相比,抗菌肽抗菌活力相对较低。即达到相同抗菌效果时,抗菌肽使用浓度可能更高[39],而高浓度的抗菌肽会增加宿主红细胞的溶血性。(3) 药物生产加工方面:与传统小分子治疗药物相比,抗菌肽序列较长,往往具有较高的生产制造成本,尤其是富含二硫化物的抗菌肽,一定程度上限制了抗菌肽在临床上的生产与使用[40–41]。
3 天然抗菌肽结构改造策略目前针对抗菌肽存在的问题现已提出许多解决策略,如对抗菌肽进行化学修饰以提高其结构稳定性、替换抗菌肽氨基酸序列提高其抗菌活性、内溶素融合抗菌肽降低其细胞毒性、设计短线性抗菌肽以降低生产成本等。
3.1 化学修饰化学修饰不仅可提高抗菌肽对蛋白酶的稳定性,改善抗菌肽药代动力学,还可以增强抗菌肽活性,如使用环化、末端/侧链修饰、D-氨基酸、与环状肽连接等[42–43]。(1) 环化,即通过连接线性抗菌肽的N-端和C-端使抗菌肽形成环状结构,是广泛提高抗菌肽稳定性的一种策略[44]。如Kamysz等通过将抗菌肽LL-37环化与线性LL-37对比发现,环状LL-37与线性LL-37抗菌活性、溶血活性相似,但稳定性更高[45];(2) 典型的抗菌肽的末端/侧链修饰有N-末端乙酰化、C-末端酰胺化等。如Li等对该实验室发现的抗菌肽L163的N-末端乙酰化后发现,N-乙酰化增强了L163对pH、血浆以及胰蛋白酶降解的稳定性[46]。(3) 改变抗菌肽手性增加抗菌肽稳定性。天然抗菌肽通常含有L-氨基酸,而非天然D-氨基酸掺入抗菌肽序列可逆转肽的立体化学,从而防止蛋白酶降解[43]。polybia CP(ILGTILGLLKSL-NH2)是一种从群居黄蜂Polybia paulista的毒液中分离的抗菌肽,研究人员通过采用D-氨基酸替代,构建了全D-氨基酸衍生的polybia CP(D-CP),发现D-CP对胰蛋白酶和糜蛋白酶降解的稳定性明显提高[47]。(4) 与环状肽(cyclotides)结合也可以提高抗菌肽稳定性。环状肽广泛分布在植物中,通常含有28−37个氨基酸和3个二硫键,其中半胱氨酸残基Cys Ⅰ–Cys Ⅳ和Cys Ⅱ–Cys Ⅴ形成梯形图案,Cys Ⅲ–Cys Ⅵ在它们之间并联形成3个保守的二硫键-环半胱氨酸结(cyclic cysteine knot,CCK),具有比线性肽更强的稳定性和生物活性,许多研究表明将抗菌肽与环状肽结合并进行相应的氨基酸修饰可以显著改善抗菌肽的药代动力学性能[48]。
3.2 氨基酸替换替换抗菌肽序列中的氨基酸可以改善抗菌肽活性低的问题。通常抗菌肽中只有少量氨基酸是抗菌活性所必需的,其他残基可以在不影响抗菌肽功能的情况下进行替换。由于苯丙氨酸(Phe)和色氨酸(Trp)等芳香族残基可促进抗菌肽形成两亲性结构,赖氨酸(Lys)等疏水性氨基酸可提高抗菌肽疏水性以提高抗菌肽抗菌活性。研究表明,在牛乳铁蛋白衍生物序列中引入芳香族氨基酸替换已有的半胱氨酸等残基,可以增加牛乳铁蛋白衍生物的抗菌活性,而不增加溶血性[42, 49]。Mourtada等发现在抗菌肽Mag(i+4)的亲水性表面用赖氨酸替换部分亲水性残基可以提高抗菌肽表面的总电荷量,抗菌活性提高[50]。另外,抗菌肽螺旋性强度与抗菌活性密切相关,大多数天然α-螺旋抗菌肽是从肽序列中的疏水性甘氨酸残基开始,因此,在抗菌肽N-末端添加甘氨酸残基或减少无规则卷曲的氨基酸残基有助于保持自身螺旋性,提高抗菌活性,例如,buforin Ⅱ (TRSSRAGLQFPVG RVHRLLRK)是一种N-末端无规卷曲而C-末端呈规则α-螺旋结构的抗菌肽,Park等研究表明,在buforin Ⅱ的N-末端删除4个氨基酸残基(TRS)后,buforin Ⅱ抗菌活性增加近2倍[51]。此外,改善抗菌肽活性也可以使用传递系统(delivery systems),即使用无机、聚合物材料、表面活性剂/脂质自组装系统以及肽自组装系统等,以增强抗菌肽的稳定性、毒性、半衰期等[43]。研究表明,通过制备抗菌肽LL-37聚乙二醇脂质体,可以增强LL-37抗菌活性、稳定性并降低红细胞溶血性[52]。
3.3 内溶素融合抗菌肽选择性融合内溶素与抗菌肽可有效改善抗菌肽对红细胞溶血的毒性。噬菌体衍生蛋白内溶素是双链DNA在复制周期结束时产生并释放的可溶性酶,用来分解细菌细胞壁的肽聚糖,从而释放病毒子代,内溶素作用机制类似于肽聚糖水解酶(peptpglycan hydrolase)[53]。内溶素通常只对噬菌体细菌宿主具有特异性杀菌活性,该特异性通常由特定的PG化学型对酶活性域(enzymatically active domain,EAD)的敏感性、细胞壁结合域(cell wall binding domain,CBD)对细胞壁或相关配体的识别来确定,从而将抗菌作用限制在特定属、种、血清型或菌株。然而,外源性应用内溶素杀菌通常仅限于细胞膜表面没有外膜或脂质的生物体(即革兰氏阳性细菌),革兰氏阴性菌由于有外膜的保护而不受内溶素的影响。但研究发现,通过将内溶素与抗菌肽选择性结合,能有效杀灭革兰氏阴性菌并减少对哺乳动物红细胞的溶血性[54]。例如由内溶素(KZ144)和靶向肽(SMAP-29,一种分离于绵羊白细胞的α-螺旋抗菌肽)共价结合而成的抗菌肽Artilysin®-Art-175在30×MIC+ 0.5 mmol/L EDTA的浓度下,6 min即可完全清除铜绿假单胞菌(Pseudomonas aeruginosa) PAO1,通过延时显微镜发现,在加入Art-175后1 min内P. aeruginosa PAO1细胞形态即发生改变,呈现不规则状。在相同浓度条件下,Art-175与P. aeruginosa PA14、P. aeruginosa PA1255持留菌共同作用1 h也能实现全部杀灭。由于抗菌肽SMAP-29对哺乳动物红细胞具有溶血性,使用L-929小鼠结缔组织成纤维细胞与Art-175共同培养以测试Art-175细胞毒性,最终发现有93%±1.4%的成纤维细胞存活,只有不足10%的细胞出现溶解或胞浆内离散颗粒,表明Art-175具有选择性毒性[55]。根据内溶素的特异性毒性,分离于P. aeruginosa的内溶素KZ144,在添加外源性膜渗透剂的情况下,应仅对P. aeruginosa有效,而我们发现,Artilysin®对革兰氏阴性菌E. coli C600的MIC为10 µg/mL,略高于colistin对E. coli C600的MIC (2 µg/mL),对革兰氏阳性菌金黄色葡萄球菌(Staphylococcus aureus) HG003的MIC为2 µg/mL,与ampicillin对S. aureus HG003的MIC一致,表现出广谱的抗菌效果。因此,通过将内溶素与抗菌肽选择性融合,在保留抗菌肽原有优势的情况下,减少细胞溶血性毒性。
3.4 设计短线性抗菌肽长序列的抗菌肽往往伴随着高昂的生产成本,可以通过设计短线性抗菌肽以降低抗菌肽成本过高的问题[41],其中序列长度不超过20个残基的短抗菌肽更容易制备且成本更低[56]。然而,研究发现,以上抗菌肽优化策略无法同时解决或改善以上所有问题。例如,虽然短线性抗菌肽具有较低的生产成本,但容易被酶水解;而氨基酸修饰可以提高抗菌肽稳定性,但生产成本更高[42]。
近来,越来越多的研究发现,基于人工智能(artificial intelligence,AI)的各种算法以学习复杂分子系统的内部相互作用为目标[57],可用于设计具有特定结构或功能要求的新分子,为合成高稳定性、低毒性以及合理成本的抗菌肽提供了良好的策略。
4 人工智能助力新型抗菌肽研发人工智能设计抗菌肽首先需建立预测模型来估计抗菌肽分子性质,以用于候选筛选,通常使用手动选择或自动学习的成分、结构、物理化学特征集来构建预测模型;候选的抗菌肽通常是通过对合理的子序列进行组合计数,然后从现有的分子库中随机选择或修改而获得。同时通过将抗菌肽氨基酸残基以字符串形式表示,以对抗菌肽氨基酸序列数据集进行机器学习/深度学习训练和分析,从而得以高效识别新型抗菌肽氨基酸序列。
4.1 机器学习用于抗菌肽研发机器学习(machine learning,ML),属于人工智能的子领域,可以从示例数据推演出基本规则,是一种从训练数据中学习的计算机决策方法,在药物开发中扮演着重要角色。ML基于定量结构-活性关系(quantitative structure activity relationship,QSAR)模型开发,QSAR模型能发现高效、稳健的计算程序,以便在数据库和虚拟库中定位具有已知活性的分子,采用人工神经网络(artificial neural networks,ANN)、支持向量机(support vector machines,SVM)、定量矩阵(quantitative matrices,QM)等方法[58],侧重考虑抗菌肽的抗菌活性、候选分子的抗菌潜力及毒性,是抗菌肽的高通量设计方法[59]。ML用于抗菌肽研发主要是利用数据驱动学习,发现和设计具有高活性的抗菌肽,这得益于对抗菌肽活性,尤其是膜活性分子的深入理解[6]。例如,Lee等利用ML构建了一个基于α-螺旋抗菌肽序列的SVM分类器,研究α-螺旋抗菌肽及其功能共性和序列同源性的相关性质。SVM用于搜索未被发现的抗菌肽序列空间,识别帕累托最优(Pareto-optimal)候选序列,同时最大化与SVM超平面的距离σ(从而最大化抗菌肽的抗菌活性)及α-螺旋的稳定性,最小化与已知抗菌肽的突变距离,继而通过杀灭分析、小角度X-射线散射(small angle X-ray scattering,SAXS)校准SVM机器学习的结果,结果表明SVM分类器能有效地呈现膜渗透所需的几何和拓扑原理,不仅可以从已知的抗菌肽中预测未知的膜活性肽,还可以识别具有多种功能的抗菌肽,并发现已有抗菌肽之间未知相互关系[60]。Boone等将遗传算法与粗糙集理论(一种透明的ML方法,用于跟踪标签的模糊性,以了解输入和输出标签之间关系的强度)相结合设计抗菌肽,这种方法的使用为抗菌肽活性序列具有和非活性序列不具有的物理化学性质提供了明确边界,据此使用监督学习边界实现抗菌肽设计,从而找到了对表皮葡萄球菌(S. epidermidis)具有高活性的抗菌肽,并发现通过该方法合成的抗菌肽具有比APD3 (https://aps.unmc.edu/AP/)数据库中的抗菌肽更易合成等特点[61]。除此之外,ML还可以与进化算法结合以改善抗菌肽红细胞溶血毒性[62]。这些结果表明,ML作为人工智能子领域,可有效预测发现或合成优化的抗菌肽。
4.2 深度学习用于抗菌肽研发随着越来越多天然抗菌肽的发现,积累了大量可用于训练的样本数据,深度学习的优势得以逐渐体现。深度学习(deep learning,DL)是ML领域中一个新的研究方向,与ML不同的是,DL的完成不需要大量的数据,具有合并自动编码(即特征生成)的优势。该方法通常将学习内容限制在具有所需属性的固定分子库中,组合低层特征形成抽象的高层表示属性类别或特征,以发现数据的分布式特征[56]。用于序列分类的常见DL包括递归神经网络(recurrent neural networks,RNN)和卷积神经网络(convolutional neural networks,CNN)[63]。如Wang等通过构建LSTM生成模型(long short-term memory,一种RNN模型)和双向LSTM分类模型,设计了具有潜在抗E. coli活性的抗菌肽短序列,其中分类模型的验证准确率为81.6%–88.9%,新型抗菌肽归类为抗菌药物的准确率高达70.6%–91.7%,这表明LSTM是寻找新型抗菌肽的有效工具[56]。中国科学院微生物研究所王军、陈义华课题组成员通过结合多种自然语言方法(natural language processing,NLP)处理LSTM、Attention以及BERT等多种神经网络模型,从人类肠道微生物组数据中识别候选的抗菌肽。NLP可以自主学习序列特征,通过识别基因组序列中的特征,甚至是低同源性的短序列来识别候选抗菌肽。通过该方法,最终发现在2 349个候选的抗菌肽序列中,化学合成的占有216个,181个具有抗菌活性,阳性率高达83%以上[64]。Sharma等基于DL的特征,采用SVM算法,构建了AniAMPpred模型以识别动物基因组中可能存在的抗菌蛋白;AniAMPpred可对不同长度的抗菌肽或非抗菌肽进行高精度分类,通过该方法在Helobdella robusta基因组中鉴定了436个可能的抗菌蛋白[65]。Das等利用自动编码器建模的分子信息潜在空间的分类器指导,提出条件潜在(属性)空间采样模型[conditional latent (attribute) space sampling,CLaSS]控制分子的产生,通过排斥采样(rejection sampling)方案生成具有所需属性的分子,并使用DL分类器和高通量分子动力学模拟中得出的物理化学特征来筛选生成的分子;在48 d内鉴定、合成测试了20种候选抗菌肽,其中有2种抗菌肽(YI12和FK13)极具治疗潜力,并且都具有较低的溶血毒性和致死性[57]。
在DL的基础上,还可以开发生成模型,用于自动从头设计具有指定特征的分子。基于DL构建的生成性深度学习(generative DL)可通过多种计算方式发现抗菌肽,它通常采用生成性对抗网络(generative adversarial networks,GAN)、可变自动编码器(variational autoencoders,VAE)或相关架构模型。GAN是通过对抗的方式,学习数据分布的生成式模型,控制生成序列的概率分布,以尽可能地覆盖具有抗菌活性的抗菌肽。如Tucs等利用GAN设计了一种抗菌肽生成模型-PepGAN,该模型可在选取活性抗菌肽和避开非活性抗菌肽取得平衡,具有良好的保真度,通过该模型设计的高活性抗菌肽,对E. coli TOP10的MIC仅为3.1 μg/mL,明显低于ampicillin对E. coli TOP10的MIC[66]。VAE作为Generative DL的模型,可以大量读取所需数据特征,通过编码、解码以及系统抽样,自动生成具有这些特征的新数据。Dean等以抗菌肽为模型,基于公开的数据库(APD3)训练Generative DL算法,生成具有抗菌活性的新抗菌肽序列,并使用VAE模型,制作潜在空间图,测量具有已知特性的抗菌肽。从而实现自动化生成新抗菌肽序列,同时预测抗菌肽活性并进行优化[67]。以上相关人工智能算法发现的新型抗菌肽及其特点如表 2所示。这些研究表明人工智能算法的使用不仅能降低发现生产有效小分子抗菌肽的成本与时间,且能优化序列降低毒性、提高活性的同时相比于天然的抗菌肽更易合成。突破了以往抗菌肽应用存在的劣势,为抗菌肽的开发与应用创造了许多机会,为实现抗菌肽的商业化生产提供了良好的前景。
Antimicrobial peptides | Research methods | Characteristics | References |
YI12, FK113 | Deep-learning classifiers and high-throughput molecular dynamics simulations | Low hemolysis in vitro and lethality in vivo | [57] |
AMP-2 | Genetic algorithm and rough set theory | Easier to synthesize than the antimicrobial peptides in the database | [61] |
GN1, GP1 | Trained RNN with data from DBAASP database to design short non-hemolytic AMPs | No hemolysis,high activity and broad-spectrum antibacterial activity | [62] |
RaCa-2 | An attentive deep learning model-AMPlify | Effective against WHO priority pathogens | [63] |
c_AMP67, c_AMP69, c_AMP660 |
Combining the neural network models (NNMs) for autonomous learning of AMP sequence features and human microbiome data resources discover AMPs | High activity against multidrug-resistant, Gram-negative bacteria and low toxicity to human cells | [64] |
AMP-6 | Peptide-specialized model based on deep learning--PepGAN | High antibacterial activity. The antibacterial effect on E. coli is stronger than ampicillin | [66] |
NN2_0050, NN2_0018 |
The LSTM model is used to understand the arrangement and frequency of amino acid residues in the peptide, so as to generate the antimicrobial peptide sequence | High activity against MDR clinical isolates, including carbapenem resistant bacteria and methicillin resistant bacteria | [68] |
GMG_01, GMG_02 |
Design novel AMP sequences through machine learning and other computational approach, based on chemophysical profiles of peptide sequences | High antibacterial activity | [69] |
5 结论与展望
抗菌肽因具有广谱的抗菌效果、不易产生耐药性以及易于合成和修饰等特点,有望成为对抗感染和微生物耐药性最有希望的候选药物之一。在未来的研究中,从多种生物中鉴定天然抗菌肽仍然是一个重要的研究热点。但是,由于天然抗菌肽存在的抗菌活性低、红细胞溶血毒性以及高合成成本等问题,限制了天然抗菌肽的应用潜力与发展前景。因此,现阶段不仅需要对已发现的天然抗菌肽进行广泛的结构-活性关系以及化学修饰研究,还需要一系列人工智能算法驱动新型抗菌肽的研发,算法理论和邻近的计算领域有望为促进新型抗菌肽的发现提供更多新途径。创造性地整合ML、DL等方法辅助抗菌肽设计与发现,在未来将使以合理的成本提供良好药代动力学的抗菌肽成为可能。
[1] | Rodríguez AA, Otero-González A, Ghattas M, Ständker L. Discovery, optimization, and clinical application of natural antimicrobial peptides. Biomedicines, 2021, 9(10): 1381. DOI:10.3390/biomedicines9101381 |
[2] | Yount NY, Bayer AS, Xiong YQ, Yeaman MR. Advances in antimicrobial peptide immunobiology. Biopolymers, 2006, 84(5): 435-458. DOI:10.1002/bip.20543 |
[3] | Teixeira V, Feio MJ, Bastos M. Role of lipids in the interaction of antimicrobial peptides with membranes. Progress in Lipid Research, 2012, 51(2): 149-177. DOI:10.1016/j.plipres.2011.12.005 |
[4] | Lazzaro BP, Zasloff M, Rolff J. Antimicrobial peptides: application informed by evolution. Science, 2020, 368(6490): eaau5480. DOI:10.1126/science.aau5480 |
[5] | Fry DE. Antimicrobial peptides. Surgical Infections, 2018, 19(8): 804-811. DOI:10.1089/sur.2018.194 |
[6] | Cardoso P, Glossop H, Meikle TG, Aburto-Medina A, Conn CE, Sarojini V, Valery C. Molecular engineering of antimicrobial peptides: microbial targets, peptide motifs and translation opportunities. Biophysical Reviews, 2021, 13(1): 35-69. DOI:10.1007/s12551-021-00784-y |
[7] | Matsuzaki K. Control of cell selectivity of antimicrobial peptides. Biochimica et Biophysica Acta: BBA--Biomembranes, 2009, 1788(8): 1687-1692. DOI:10.1016/j.bbamem.2008.09.013 |
[8] | Gan BH, Gaynord J, Rowe SM, Deingruber T, Spring DR. The multifaceted nature of antimicrobial peptides: current synthetic chemistry approaches and future directions. Chemical Society Reviews, 2021, 50(13): 7820-7880. DOI:10.1039/D0CS00729C |
[9] | Brogden KA. Antimicrobial peptides: pore formers or metabolic inhibitors in bacteria?. Nature Reviews Microbiology, 2005, 3(3): 238-250. DOI:10.1038/nrmicro1098 |
[10] | Zhang QY, Yan ZB, Meng YM, Hong XY, Shao G, Ma JJ, Cheng XR, Liu J, Kang J, Fu CY. Antimicrobial peptides: mechanism of action, activity and clinical potential. Military Medical Research, 2021, 8(1): 48. DOI:10.1186/s40779-021-00343-2 |
[11] | Zhang YM, Rock CO. Membrane lipid homeostasis in bacteria. Nature Reviews Microbiology, 2008, 6(3): 222-233. DOI:10.1038/nrmicro1839 |
[12] | Matsuzaki K. Why and how are peptide-lipid interactions utilized for self-defense? Magainins and tachyplesins as archetypes. Biochimica et Biophysica Acta: BBA--Biomembranes, 1999, 1462(1/2): 1-10. |
[13] | Enoki TA, Moreira-Silva I, Lorenzon EN, Cilli EM, Perez KR, Riske KA, Lamy MT. Antimicrobial peptide K0-W6-Hya1 induces stable structurally modified lipid domains in anionic membranes. Langmuir, 2018, 34(5): 2014-2025. DOI:10.1021/acs.langmuir.7b03408 |
[14] | Guilhelmelli F, Vilela N, Albuquerque P, Derengowski Lda S, Silva-Pereira I, Kyaw CM. Antibiotic development challenges: the various mechanisms of action of antimicrobial peptides and of bacterial resistance. Frontiers in Microbiology, 2013, 4: 353. |
[15] | Andersson DI, Hughes D, Kubicek-Sutherland JZ. Mechanisms and consequences of bacterial resistance to antimicrobial peptides. Drug Resistance Updates, 2016, 26: 43-57. DOI:10.1016/j.drup.2016.04.002 |
[16] | Epand RM, Walker C, Epand RF, Magarvey NA. Molecular mechanisms of membrane targeting antibiotics. Biochimica et Biophysica Acta: BBA--Biomembranes, 2016, 1858(5): 980-987. DOI:10.1016/j.bbamem.2015.10.018 |
[17] |
Zhang X, Gong L. Antimicrobial mechanism of antimicrobial peptide and research progress. Chinese Journal of Tissue Engineering Research, 2020, 24(10): 1634-1640.
(in Chinese) 张溪, 弓磊. 抗菌肽抗菌机制及研究热点. 中国组织工程研究, 2020, 24(10): 1634-1640. DOI:10.3969/j.issn.2095-4344.2202 |
[18] | Wang JJ, Dou XJ, Song J, Lyu YF, Zhu X, Xu L, Li WZ, Shan AS. Antimicrobial peptides: promising alternatives in the post feeding antibiotic era. Medicinal Research Reviews, 2019, 39(3): 831-859. DOI:10.1002/med.21542 |
[19] | Pálffy R, Gardlík R, Behuliak M, Kadasi L, Turna J, Celec P. On the physiology and pathophysiology of antimicrobial peptides. Molecular Medicine: Cambridge, Mass, 2009, 15(1/2): 51-59. |
[20] | Perrin BS Jr, Pastor RW. Simulations of membrane-disrupting peptides I: alamethicin pore stability and spontaneous insertion. Biophysical Journal, 2016, 111(6): 1248-1257. DOI:10.1016/j.bpj.2016.08.014 |
[21] | Ramamoorthy A, Lee DK, Narasimhaswamy T, Nanga RPR. Cholesterol reduces pardaxin's dynamics—a barrel-stave mechanism of membrane disruption investigated by solid-state NMR. Biochimica et Biophysica Acta: BBA--Biomembranes, 2010, 1798(2): 223-227. DOI:10.1016/j.bbamem.2009.08.012 |
[22] | Pieta P, Majewska M, Su ZF, Grossutti M, Wladyka B, Piejko M, Lipkowski J, Mak P. Physicochemical studies on orientation and conformation of a new bacteriocin BacSp222 in a planar phospholipid bilayer. Langmuir: the ACS Journal of Surfaces and Colloids, 2016, 32(22): 5653-5662. DOI:10.1021/acs.langmuir.5b04741 |
[23] | Moll GN, Konings WN, Driessen AJ. Bacteriocins: mechanism of membrane insertion and pore formation. Antonie Van Leeuwenhoek, 1999, 76(1/2/3/4): 185-198. |
[24] | Fernandez DI, Le Brun AP, Whitwell TC, Sani MA, James M, Separovic F. The antimicrobial peptide aurein 1.2 disrupts model membranes via the carpet mechanism. Physical Chemistry Chemical Physics: PCCP, 2012, 14(45): 15739-15751. DOI:10.1039/c2cp43099a |
[25] | Majewska M, Zamlynny V, Pieta IS, Nowakowski R, Pieta P. Interaction of LL-37 human cathelicidin peptide with a model microbial-like lipid membrane. Bioelectrochemistry, 2021, 141: 107842. DOI:10.1016/j.bioelechem.2021.107842 |
[26] | Matsuzaki K. Membrane permeabilization mechanisms. Advances in Experimental Medicine and Biology. Singapore: Springer Singapore, 2019: 9–16. |
[27] | Bin Hafeez A, Jiang XK, Bergen PJ, Zhu Y. Antimicrobial peptides: an update on classifications and databases. International Journal of Molecular Sciences, 2021, 22(21): 11691. DOI:10.3390/ijms222111691 |
[28] | Cheng JTJ, Hale JD, Elliot M, Hancock REW, Straus SK. Effect of membrane composition on antimicrobial peptides aurein 2.2 and 2. 3 from Australian southern bell frogs. Biophysical Journal, 2009, 96(2): 552-565. DOI:10.1016/j.bpj.2008.10.012 |
[29] | Guha S, Ferrie RP, Ghimire J, Ventura CR, Wu E, Sun LS, Kim SY, Wiedman GR, Hristova K, Wimley WC. Applications and evolution of melittin, the quintessential membrane active peptide. Biochemical Pharmacology, 2021, 193: 114769. DOI:10.1016/j.bcp.2021.114769 |
[30] | Müller A, Ulm H, Reder-Christ K, Sahl HG, Schneider T. Interaction of type A lantibiotics with undecaprenol-bound cell envelope precursors. Microbial Drug Resistance: Larchmont, N Y, 2012, 18(3): 261-270. DOI:10.1089/mdr.2011.0242 |
[31] | Haney EF, Petersen AP, Lau CK, Jing WG, Storey DG, Vogel HJ. Mechanism of action of puroindoline derived tryptophan-rich antimicrobial peptides. Biochimica et Biophysica Acta: BBA--Biomembranes, 2013, 1828(8): 1802-1813. DOI:10.1016/j.bbamem.2013.03.023 |
[32] | Scocchi M, Mardirossian M, Runti G, Benincasa M. Non-membrane permeabilizing modes of action of antimicrobial peptides on bacteria. Current Topics in Medicinal Chemistry, 2016, 16(1): 76-88. |
[33] |
Liu XL, Yang WX, Ma YL, Wang D, Chen H. Research progress in the formation mechanism and treatment of persister. Chinese Journal of Infection Control, 2020, 19(2): 184-188.
(in Chinese) 刘小龙, 杨万霞, 马延龄, 王丹, 陈昊. 持留菌形成机制及治疗研究进展. 中国感染控制杂志, 2020, 19(2): 184-188. |
[34] | Defraine V, Schuermans J, Grymonprez B, Govers SK, Aertsen A, Fauvart M, Michiels J, Lavigne R, Briers Y. Efficacy of artilysin art-175 against resistant and persistent Acinetobacter baumannii. Antimicrobial Agents and Chemotherapy, 2016, 60(6): 3480-3488. DOI:10.1128/AAC.00285-16 |
[35] | Yeung ATY, Gellatly SL, Hancock REW. Multifunctional cationic host defence peptides and their clinical applications. Cellular and Molecular Life Sciences: CMLS, 2011, 68(13): 2161-2176. DOI:10.1007/s00018-011-0710-x |
[36] | Silveira RF, Roque-Borda CA, Vicente EF. Antimicrobial peptides as a feed additive alternative to animal production, food safety and public health implications: an overview. Animal Nutrition, 2021, 7(3): 896-904. DOI:10.1016/j.aninu.2021.01.004 |
[37] | Fasina YO, Obanla T, Dosu G, Muzquiz S. Significance of endogenous antimicrobial peptides on the health of food animals. Frontiers in Veterinary Science, 2021, 8: 585266. DOI:10.3389/fvets.2021.585266 |
[38] | Żyrek D, Wajda A, Czechowicz P, Nowicka J, Jaśkiewicz M, Neubauer D, Kamysz W. The antimicrobial activity of omiganan alone and in combination against Candida isolated from vulvovaginal candidiasis and bloodstream infections. Antibiotics: Basel, Switzerland, 2021, 10(8): 1001. |
[39] | Chen X, Zhang M, Zhou CH, Kallenbach NR, Ren DC. Control of bacterial persister cells by Trp/Arg-containing antimicrobial peptides. Applied and Environmental Microbiology, 2011, 77(14): 4878-4885. DOI:10.1128/AEM.02440-10 |
[40] | Mahlapuu M, Håkansson J, Ringstad L, Björn C. Antimicrobial peptides: an emerging category of therapeutic agents. Frontiers in Cellular and Infection Microbiology, 2016, 6: 194. |
[41] | Lau QY, Li JG, Sani MA, Sinha S, Li Y, Ng FM, Kang CB, Bhattacharjya S, Separovic F, Verma C, Chia CSB. Elucidating the bactericidal mechanism of action of the linear antimicrobial tetrapeptide BRBR-NH2. Biochimica et Biophysica Acta: BBA--Biomembranes, 2018, 1860(8): 1517-1527. DOI:10.1016/j.bbamem.2018.05.004 |
[42] | Han YJ, Zhang ML, Lai R, Zhang ZY. Chemical modifications to increase the therapeutic potential of antimicrobial peptides. Peptides, 2021, 146: 170666. DOI:10.1016/j.peptides.2021.170666 |
[43] | Kumar P, Kizhakkedathu JN, Straus SK. Antimicrobial peptides: diversity, mechanism of action and strategies to improve the activity and biocompatibility in vivo. Biomolecules, 2018, 8(1): 4. DOI:10.3390/biom8010004 |
[44] | Gunasekera S, Muhammad T, Strömstedt AA, Rosengren KJ, Göransson U. Backbone cyclization and dimerization of LL-37-derived peptides enhance antimicrobial activity and proteolytic stability. Frontiers in Microbiology, 2020, 11: 168. DOI:10.3389/fmicb.2020.00168 |
[45] | Kamysz E, Sikorska E, Karafova A, Dawgul M. Synthesis, biological activity and conformational analysis of head-to-tail cyclic analogues of LL37 and histatin 5. Journal of Peptide Science, 2012, 18(9): 560-566. DOI:10.1002/psc.2434 |
[46] | Li DD, Yang YH, Li RF, Huang L, Wang ZC, Deng QW, Dong SB. N-terminal acetylation of antimicrobial peptide L163 improves its stability against protease degradation. Journal of Peptide Science: an Official Publication of the European Peptide Society, 2021, 27(9): e3337. |
[47] | Jia F, Wang J, Peng J, Zhao P, Kong Z, Wang K, Yan W, Wang R. D-amino acid substitution enhances the stability of antimicrobial peptide polybia-CP. Acta Biochimica et Biophysica Sinica, 2017, 49(10): 916-925. DOI:10.1093/abbs/gmx091 |
[48] | Srivastava S, Dashora K, Ameta KL, Singh NP, El-Enshasy HA, Pagano MC, Hesham AEL, Sharma GD, Sharma M, Bhargava A. Cysteine-rich antimicrobial peptides from plants: the future of antimicrobial therapy. Phytotherapy Research: PTR, 2021, 35(1): 256-277. DOI:10.1002/ptr.6823 |
[49] | Haug BE, Skar ML, Svendsen JS. Bulky aromatic amino acids increase the antibacterial activity of 15-residue bovine lactoferricin derivatives. Journal of Peptide Science: an Official Publication of the European Peptide Society, 2001, 7(8): 425-432. |
[50] | Mourtada R, Herce HD, Yin DJ, Moroco JA, Wales TE, Engen JR, Walensky LD. Design of stapled antimicrobial peptides that are stable, nontoxic and kill antibiotic-resistant bacteria in mice. Nature Biotechnology, 2019, 37(10): 1186-1197. DOI:10.1038/s41587-019-0222-z |
[51] | Park CB, Yi KS, Matsuzaki K, Kim MS, Kim SC. Structure-activity analysis of buforin II, a histone H2A-derived antimicrobial peptide: the proline hinge is responsible for the cell-penetrating ability of buforin II. Proceedings of the National Academy of Sciences of the United States of America, 2000, 97(15): 8245-8250. DOI:10.1073/pnas.150518097 |
[52] | Ron-Doitch S, Sawodny B, Kühbacher A, David MMN, Samanta A, Phopase J, Burger-Kentischer A, Griffith M, Golomb G, Rupp S. Reduced cytotoxicity and enhanced bioactivity of cationic antimicrobial peptides liposomes in cell cultures and 3D epidermis model against HSV. Journal of Controlled Release, 2016, 229: 163-171. DOI:10.1016/j.jconrel.2016.03.025 |
[53] | Vázquez-López R, Solano-Gálvez SG, Juárez Vignon-Whaley JJ, Abello Vaamonde JA, Padró Alonzo LA, Rivera Reséndiz A, Muleiro Álvarez M, Vega López EN, Franyuti-Kelly G, Álvarez-Hernández DA, Moncaleano Guzmán V, Juárez Bañuelos JE, Marcos Felix J, González Barrios JA, Barrientos Fortes T. Acinetobacter baumannii resistance: a real challenge for clinicians. Antibiotics: Basel, Switzerland, 2020, 9(4): 205. |
[54] | Antonova NP, Vasina DV, Rubalsky EO, Fursov MV, Savinova AS, Grigoriev IV, Usachev EV, Shevlyagina NV, Zhukhovitsky VG, Balabanyan VU, Potapov VD, Aleshkin AV, Makarov VV, Yudin SM, Gintsburg AL, Tkachuk AP, Gushchin VA. Modulation of endolysin LysECD7 bactericidal activity by different peptide tag fusion. Biomolecules, 2020, 10(3): 440. DOI:10.3390/biom10030440 |
[55] | Briers Y, Walmagh M, Grymonprez B, Biebl M, Pirnay JP, Defraine V, Michiels J, Cenens W, Aertsen A, Miller S, Lavigne R. Art-175 is a highly efficient antibacterial against multidrug-resistant strains and persisters of Pseudomonas aeruginosa. Antimicrobial Agents and Chemotherapy, 2014, 58(7): 3774-3784. DOI:10.1128/AAC.02668-14 |
[56] | Wang C, Garlick S, Zloh M. Deep learning for novel antimicrobial peptide design. Biomolecules, 2021, 11(3): 471. DOI:10.3390/biom11030471 |
[57] | Das P, Sercu T, Wadhawan K, Padhi I, Gehrmann S, Cipcigan F, Chenthamarakshan V, Strobelt H, dos Santos C, Chen PY, Yang YY, Tan JPK, Hedrick J, Crain J, Mojsilovic A. Accelerated antimicrobial discovery via deep generative models and molecular dynamics simulations. Nature Biomedical Engineering, 2021, 5(6): 613-623. DOI:10.1038/s41551-021-00689-x |
[58] | Lee MW, Lee EY, Ferguson AL, Wong GCL. Machine learning antimicrobial peptide sequences: some surprising variations on the theme of amphiphilic assembly. Current Opinion in Colloid & Interface Science, 2018, 38: 204-213. |
[59] | Cardoso MH, Orozco RQ, Rezende SB, Rodrigues G, Oshiro KGN, Cândido ES, Franco OL. Computer-aided design of antimicrobial peptides: are we generating effective drug candidates?. Frontiers in Microbiology, 2020, 10: 3097. DOI:10.3389/fmicb.2019.03097 |
[60] | Lee EY, Fulan BM, Wong GCL, Ferguson AL. Mapping membrane activity in undiscovered peptide sequence space using machine learning. Proceedings of the National Academy of Sciences of the United States of America, 2016, 113(48): 13588-13593. DOI:10.1073/pnas.1609893113 |
[61] | Boone K, Wisdom C, Camarda K, Spencer P, Tamerler C. Combining genetic algorithm with machine learning strategies for designing potent antimicrobial peptides. BMC Bioinformatics, 2021, 22(1): 239. DOI:10.1186/s12859-021-04156-x |
[62] | Capecchi A, Cai XG, Personne H, Köhler T, Van Delden C, Reymond JL. Machine learning designs non-hemolytic antimicrobial peptides. Chemical Science, 2021, 12(26): 9221-9232. DOI:10.1039/D1SC01713F |
[63] | Li CK, Sutherland D, Hammond SA, Yang C, Taho F, Bergman L, Houston S, Warren RL, Wong T, Hoang LMN, Cameron CE, Helbing CC, Birol I. AMPlify: attentive deep learning model for discovery of novel antimicrobial peptides effective against WHO priority pathogens. BMC Genomics, 2022, 23(1): 77. DOI:10.1186/s12864-022-08310-4 |
[64] | Ma Y, Guo ZY, Xia BB, Zhang YW, Liu XL, Yu Y, Tang N, Tong XM, Wang M, Ye X, Feng J, Chen YH, Wang J. Identification of antimicrobial peptides from the human gut microbiome using deep learning. Nature Biotechnology, 2022, 40(6): 921-931. DOI:10.1038/s41587-022-01226-0 |
[65] | Sharma R, Shrivastava S, Kumar Singh S, Kumar A, Saxena S, Kumar Singh R. AniAMPpred: artificial intelligence guided discovery of novel antimicrobial peptides in animal kingdom. Briefings in Bioinformatics, 2021, 22(6): bbab242. DOI:10.1093/bib/bbab242 |
[66] | Tucs A, Tran DP, Yumoto A, Ito Y, Uzawa T, Tsuda K. Generating ampicillin-level antimicrobial peptides with activity-aware generative adversarial networks. ACS Omega, 2020, 5(36): 22847-22851. DOI:10.1021/acsomega.0c02088 |
[67] | Dean SN, Walper SA. Variational autoencoder for generation of antimicrobial peptides. ACS Omega, 2020, 5(33): 20746-20754. DOI:10.1021/acsomega.0c00442 |
[68] | Nagarajan D, Nagarajan T, Roy N, Kulkarni O, Ravichandran S, Mishra M, Chakravortty D, Chandra N. Computational antimicrobial peptide design and evaluation against multidrug-resistant clinical isolates of bacteria. Journal of Biological Chemistry, 2018, 293(10): 3492-3509. DOI:10.1074/jbc.M117.805499 |
[69] | Maccari G, Di Luca M, Nifosí R, Cardarelli F, Signore G, Boccardi C, Bifone A. Antimicrobial peptides design by evolutionary multiobjective optimization. PLoS Computational Biology, 2013, 9(9): e1003212. DOI:10.1371/journal.pcbi.1003212 |