HLAEquity: Examining biases in pan-allele peptide-HLA binding predictors.
HLAEquity: Examining biases in pan-allele peptide-HLA binding predictors.
复制标题
HLA公平性：探究泛等位基因肽 - HLA结合预测因子中的偏差
DOI：
10.1016/j.isci.2023.108613
复制
发表时间：
2024-01-19
期刊：
影响因子：
5.8
通讯作者：
Kavraki, Lydia E.
中科院分区：
文献类型：
--
作者：
Conev, Anja;Fasoulis, Romanos;Hall-Swan, Sarah;Ferreira, Rodrigo;Kavraki, Lydia E.
肽 - 人类白细胞抗原（pHLA）结合预测在筛选个性化肽疫苗的候选肽方面至关重要。机器学习（ML）pHLA结合预测工具通过大量数据进行训练，在筛选候选肽方面颇为有效。大多数机器学习模型宣称能够推广应用于训练过程中未见过的人类白细胞抗原（HLA）等位基因（即 “泛等位基因” 模型）。然而，使用等位基因含量不均衡的数据集引发了对模型性能存在偏差的担忧。首先，我们研究了两种基于机器学习的泛等位基因pHLA结合预测工具的数据偏差。我们发现，pHLA数据集过度呈现了来自高收入国家地理人群的等位基因。其次，我们表明已识别出的数据偏差在机器学习模型中持续存在，导致算法偏差，并且对于低收入地理人群中表达的等位基因，模型性能欠佳。我们提请注意这种偏差可能产生的治疗后果，并对使用 “泛等位基因” 这一术语来描述用当前可用公共数据集训练的模型提出质疑。
pHLA结合数据中，高收入人群中常见的HLA比低收入人群中的更多
训练数据中的HLA偏差影响泛等位基因模型的性能
泛等位基因预测工具对训练数据集中未出现的HLA的预测准确性较低
泛等位基因预测工具对低收入人群中表达的HLA的预测准确性较低
免疫系统；计算生物信息学；机器学习；人文地理学
Peptide-HLA (pHLA) binding prediction is essential in screening peptide candidates for personalized peptide vaccines. Machine learning (ML) pHLA binding prediction tools are trained on vast amounts of data and are effective in screening peptide candidates. Most ML models report the ability to generalize to HLA alleles unseen during training ("pan-allele" models). However, the use of datasets with imbalanced allele content raises concerns about biased model performance. First, we examine the data bias of two ML-based pan-allele pHLA binding predictors. We find that the pHLA datasets overrepresent alleles from geographic populations of high-income countries. Second, we show that the identified data bias is perpetuated within ML models, leading to algorithmic bias and subpar performance for alleles expressed in low-income geographic populations. We draw attention to the potential therapeutic consequences of this bias, and we challenge the use of the term “pan-allele” to describe models trained with currently available public datasets. PHLA binding data have HLAs more common in high-income populations than low-income ones HLA bias in training data affects the pan-allele models’ performance Pan-allele predictors have lower accuracy for HLAs not found in training datasets Pan-allele predictors have lower accuracy for HLAs expressed in lower income populations Immune system; Computational bioinformatics; Machine learning; Human Geography
登录
查看更多内容
影响因子：
14.9
作者：
Barker, Dominic J.;Maccari, Giuseppe;Georgiou, Xenia;Cooper, Michael A.;Flicek, Paul;Robinson, James;Marsh, Steven G. E.
通讯作者：
Marsh, Steven G. E.
DOI：
10.1016/j.patter.2021.100347
发表时间：
2021-10-08
期刊：
Patterns (New York, N.Y.)
影响因子：
--
作者：
Norori N;Hu Q;Aellen FM;Faraci FD;Tzovara A
通讯作者：
Tzovara A
影响因子：
23.8
作者：
Chu, Yanyi;Zhang, Yan;Wei, Dong-Qing
通讯作者：
Wei, Dong-Qing
影响因子：
3.7
作者：
Nielsen M;Lundegaard C;Blicher T;Lamberth K;Harndahl M;Justesen S;Røder G;Peters B;Sette A;Lund O;Buus S
通讯作者：
Buus S
影响因子：
9.3
作者：
O'Donnell, Timothy J.;Rubinsteyn, Alex;Laserson, Uri
通讯作者：
Laserson, Uri