Table 1 The summary of neoantigen prediction software

HLAminer [52]

Based on the shotgun sequencing database from Illumina platform, the HLA type was predicted by orienting the assembly of shotgun sequence data and comparing it with the reference allele sequence database

VariantEffect Predictor Tool [53]

Automate annotations in a standard way to reduce manual review time, annotate and prioritize variants

NetMHCpan [54]

Sequence comparison method based on artificial neural network, and predict the affinity of peptide-MHC-I type molecular

UCSC browser [55]

Based on sequence search, the fusion of multiple databases can provide fast and accurate access to any gene segment

CloudNeo pipeline [56]

Docker container was used to complete the tasks in the workflow. After the mutant VCF file and bam file representing HLA typing were input respectively, the HLA affinity prediction of all mutant peptides was obtained

OptiType [57]

The HLA typing algorithm based on integer linear programming provides sequencing databases including RNA, exome and whole genome

ATHLATES [58]

Assembly, allele recognition and allele pair inference were applied to short sequences, and the HLA genotyping at allele level was achieved by exon sequencing

pVAC-Seq [59]

To integrate tumor mutation and expression data and identify personalized mutagens by tumor sequencing

MuPeXI [60]

The extraction and induction of mutant peptides can roughly identify tumor-specific peptides, predict their immunogenicity, and evaluate their potential for new epitopes

Strelka [61]

Based on a new Bayesian model, the matching tumor-normal sample sequencing data was used to analyze and predict somatic cell variation, with high accuracy and sensitivity

Strelka2 [62]

Based on the mixed model, the error parameters of each sample insertion or deletion were estimated, and the liquid tumor analysis was improved

VarScan2 [63]

Somatic and copy number mutations in tumor-normal exome data were detected by heuristic statistical algorithm

Somaticseq [64]

Based on a randomized enhancement algorithm, more than 70 individual genome and sequencing features were extracted for each candidate site to accurately detect somatic mutations

SMMPMBEC [65]

Using matrix as a Bayesian prior, based on the optimal neural network predicting peptide with MHC-I type molecules

NeoPredPipe [66]

Based on a pipeline connecting commonly used bioinformatic software via custom python scripts to provide neoantigen burden, tumor heterogeneity, immune stimulation potential and HLA haplotype of patients