CN107704727A - Neoantigen Activity Prediction and sort method based on tumour neoantigen characteristic value - Google Patents

Abstract

The invention discloses a neoantigen immune activity scoring and sorting method based on tumor neoantigen characteristic values, comprising the following steps: input of WGS/WES and RNA-seq sequencing data of tumor-normal samples, prediction of tumor somatic mutation and Annotation, calculation of related eigenvalues; prediction of MHC‑I combined with neoantigens based on tumor somatic mutations, calculation of related eigenvalues; extraction of related eigenvalues of neoantigens; setting of scoring functions for neoantigen activity; scoring functions based on neoantigen activity Neoantigen sequencing. The present invention first analyzes and calculates the mutation of tumor somatic cells and completes the mutation annotation, calculates some characteristic values, then predicts MHC-I binding neoantigens, and calculates some characteristic values; extracts all relevant characteristic values of tumor neoantigens, and then sets The neoantigen activity scoring function, and finally the neoantigens are sorted by the neoantigen activity scoring function. Compared with traditional screening methods, this method is more efficient and accurate, and has important application value for tumor immunotherapy.

Description

Neoantigen Activity Prediction and Ranking Method Based on Tumor Neoantigen Characteristic Value

technical field

The present invention relates to the field of tumor immunotherapy, in particular to a neoantigen activity scoring and sorting method based on characteristic values of tumor neoantigens.

Background technique

In recent years, tumor immunotherapy has shined brilliantly, clinical trials have continuously achieved breakthroughs, and the cure rate and effective remission rate have continued to increase. Efficient and precise screening of tumor neoantigens is an extremely important and basic work in tumor immunotherapy, especially for tumor immunotherapy such as TCR-T/TIL and individualized vaccines.

At present, the current scheme for screening tumor neoantigens is two steps: step 1, based on WGS/WES data of tumor-normal tissue, using tools such as Mutect/Varscan to calculate the gene mutation of tumor cells; step 2, calling NetMHCpan et al. algorithms predict MHC-I binding neoantigens.

At present, there is no effective method to sort antigens based on their activity to improve the efficiency of antigen screening. In the above scheme, since the activity of the predicted MHC-I binding neoantigens is not sorted, it will bring a huge workload to the experimental verification, resulting in low screening efficiency of tumor neoantigens.

Contents of the invention

Aiming at the deficiencies in the above-mentioned prior art, the present invention provides a neoantigen activity scoring and sorting method based on the characteristic value of tumor neoantigens, which can greatly reduce the workload of experimental verification, and further realize efficient and accurate screening of tumor neoantigens .

Technical scheme of the present invention is realized like this:

The neoantigen immune activity prediction and sorting method based on tumor neoantigen characteristic value is characterized in that it comprises the following steps:

(1), input of WGS/WES and RNA-SEQ sequencing data of tumor-normal samples: input whole-genome sequencing data WGS or whole exome sequencing data WES, transcriptome sequencing data RNA-SEQ of tumor-normal samples;

(2) Prediction and annotation of tumor somatic mutations, and calculation of related feature values: Based on the sequencing data input in step (1), use Varscan or Mutect tool to analyze and calculate tumor somatic mutations, and call VEP (Variant Effect Prediction) tool to complete Mutation annotation, call PyClone, Kallisto, Varscan or Mutect tools to calculate the following feature values: mutant gene clone ratio, mutant gene expression value TPM, allelic mutation frequency VAF;

(3) MHC-I binding neoantigen prediction based on tumor somatic mutations and calculation of related feature values: Based on the tumor somatic mutations and annotation data in step (2), use NetMHCpan, Netchop, and OptiType tools to predict MHC-I binding Neoantigen, and calculate the following eigenvalues: mutated peptides and MHC affinity ranking percentage, unmutated peptides and MHC affinity ranking percentage, peptide clipping presentation efficiency;

(4) Extraction of all relevant eigenvalues of neoantigens: for the predicted MHC-I binding neoantigens in step (3), extract all relevant eigenvalues of tumor neoantigens;

(5) Setting of neoantigen activity scoring function: for the neoantigen characteristic value extracted in step (4), set the neoantigen activity scoring function;

(6) Neoantigen sorting based on the neoantigen activity scoring function: the neoantigens are sorted by the neoantigen activity scoring function.

As a preference, in step (4), the neoantigen-related feature values include R _m , A, R _n , E, NC, CL, wherein:

R _m —the percentage of mutant peptides and MHC affinity ranking, calculated by NetMHCpan;

A—allele mutation frequency VAF, calculated by Varscan/Mutect/Strelka2;

R _n —the percentage of unmutated peptides and MHC affinity ranking, calculated by NetMHCpan;

E—mutant gene expression value TPM, calculated by Kallisto;

NC—peptide clipping and presentation efficiency, calculated by netchop;

CL—mutant gene clone ratio, calculated by pyclone.

As a preference, in step (5), the proposed neoantigen activity prediction scoring function is:

Neo_Score = abundance dissimilarity clonality;

clonality = NC · CL;

abundance=L(Rm)·A·tanh(E/k);

dissimilarity=(1-L(Rn)/2));

Wherein: L(x)=1/(1+e ^5(x-2) ), tanh(x) is hyperbolic tangent function;

k is the transcript expression abundance threshold, the default value is 1.

As a preference, in step (6), the algorithm process of sorting the neoantigens through the neoantigen activity prediction function is as follows:

a), for all predicted MHC-I binding neoantigens, call the neoantigen activity prediction function Neo_score to calculate the predicted value of neoantigen activity;

b), based on the predicted value of neoantigen activity, the neoantigens are sorted using a quick sort algorithm;

c), output the neoantigen sorting results.

Design idea and beneficial effect of the present invention that have adopted above-mentioned technical scheme are:

The technical solution of the present invention proposes a scoring function for tumor neoantigen activity prediction, which sorts the predicted activity of MHC-I binding neoantigens based on the neoantigen activity prediction function, thereby realizing efficient and accurate tumor neoantigen screening.

This function is designed based on the complete process of tumor neoantigen generation, cleavage and transport, and the combination of neoantigen and MHC. The scoring function is divided into three parts. Among them, clonality measures the efficiency of neoantigens from mutations to short peptides NC and neoantigens The proportion of CL distributed in all tumor cells is an important factor affecting the efficacy of tumor vaccines; abundance (abundance) measures the expression of neoantigens and the efficiency of neoantigens combining with MHC-I to form pMHC complexes, and the expression of neoantigen mutant genes The higher the amount E, the higher the allelic mutation frequency A, the higher the binding affinity R _m between the complexes (the smaller the IC50 value), and the stronger the immunogenicity; the dissimilarity measures the difference between the mutant peptide and the corresponding normal The difference in peptide affinity R _{n is} due to the existence of the central tolerance mechanism, the greater the difference between the two, the stronger the specificity of the neoantigen, and the smaller the side effects of the treatment. The two mapping functions in the function are used to normalize (0 to 1) calculated values, the threshold 2 in L(x) is the screening threshold for peptide-MHC binding affinity, and tanh(x/k) ensures that the expression abundance of neoantigens exceeds the set value. When the threshold k is set, the change of the function value tends to be gentle.

This activity prediction function takes into account the comprehensive influencing factors in the process of neoantigen production, and the sorted antigens are more meaningful and applicable.

Description of drawings

Fig. 1 is a schematic diagram of the neoantigen activity method and sorting method based on tumor neoantigen characteristic values described in the embodiment of the present invention.

Detailed ways

The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described here are only used to illustrate and explain the present invention, and are not intended to limit the present invention.

Embodiment one:

As shown in Figure 1, a neoantigen activity method and sorting method based on tumor neoantigen characteristic value comprises the following steps:

Step 101: Input of WGS/WES and RNA-seq sequencing data of tumor-normal samples (using melanoma patient sample-mel_21, Science 2015: Carreno B M, Magrini V, Beckerhapak M, et al.Cancerimmunotherapy.A dendritic cell vaccine increases the breadth and diversity of melanoma neoantigen-specific T cells.[J].Science,2015,348(6236):803-8.)

Step 102: Prediction and annotation of tumor somatic mutations, and calculation of related feature values: based on WGS/WES and RNA-seq sequencing data of tumor-normal samples, use tools such as Varscan/Mutect to analyze and calculate tumor somatic mutations, and call VEP The (Variant Effect Prediction) tool completes mutation annotation, and calls PyClone, Kallisto, and Varscan/Mutect tools to calculate the following feature values: mutant gene clone ratio CL, mutant gene expression value TPM, and allelic mutation frequency VAF. Taking the eigenvalues corresponding to the active peptides in the literature as an example, the calculated E(TPM) and A(VAF) values for the three peptides of the patient sample mel_21 are as follows:

Table 1

Step 103: Prediction of MHC-I binding neoantigens based on tumor somatic mutations and calculation of related feature values: Based on the tumor somatic mutations and annotation data in step (1), call NetMHCpan, Netchop, OptiType tools to predict MHC-I binding Neoantigens, and the following characteristic values were calculated: affinity ranking percentage between mutated peptides and MHC R _m , affinity ranking percentage between unmutated peptides and MHC R _n , and peptide cleavage and presentation efficiency NC. Taking the eigenvalues corresponding to the active peptides in the literature as an example, the calculated values of R _m , R _n and NC for the three peptides of the patient sample mel_21 are as follows:

Table 2

Step 104: Extraction of neoantigen-related eigenvalues: For the MHC-I binding neoantigens predicted in step 103, extract all relevant eigenvalues of tumor neoantigens, and the neoantigens and eigenvalues of patient sample 1 mel_21 are shown in Table 1 and Table 1. 2;

Step 105: setting of neoantigen activity scoring function: setting neoantigen activity scoring function for the neoantigen characteristic value extracted in step 104;

Step 106: Neoantigen sorting based on the neoantigen activity scoring function: sort the neoantigens by the neoantigen activity scoring function, and Table 3 shows the scoring (Neo_Score) of the neoantigens verified by the PMHC activity experiment in the patient sample mel_21 and The ranking (Rank) in the corpus.

table 3

Among the three identified neoantigens, CLNEYHLFL showed the immune activity of activating CD8+ T cells before stimulating DC cells with tumor vaccines, and the other two had different degrees of immune activity after using vaccines to enhance the immune system. immune activity. We can see that KMIGNHLWV and CLNEYHLFL are ranked Top 3 (the total number of candidate neoantigens is 94). However, AMFWSVPTS is ranked 33rd in our ranking results due to its low expression level in both in vitro experiments and human tumor microenvironment, which determines its weak immunogenicity and immune response. The experimental results are quite consistent with our predicted ranking. .

Embodiment two:

As shown in Figure 1, a neoantigen activity method and sorting method based on tumor neoantigen characteristic value comprises the following steps:

Step 101: Input of WGS/WES and RNA-seq sequencing data of tumor-normal samples (using melanoma patient sample 2 mel_38, Science 2015: Carreno B M, Magrini V, Beckerhapak M, et al.Cancerimmunotherapy.A dendritic cell vaccine increases the breadth and diversity of melanoma neoantigen-specific T cells.[J].Science,2015,348(6236):803-8.)

Step 102: Prediction and annotation of tumor somatic mutations, and calculation of related feature values: based on WGS/WES and RNA-seq sequencing data of tumor-normal samples, use tools such as Varscan/Mutect to analyze and calculate tumor somatic mutations, and call VEP (Variant Effect Prediction) tool completes the mutation annotation, calls PyClone, Kallisto, Varscan/Mutect tools to calculate the following eigenvalues: mutant gene clone ratio CL, mutant gene expression value TPM, allelic mutation frequency VAF; active peptides in the literature Take the eigenvalues corresponding to the segment as an example, the E(TPM) and A(VAF) values calculated corresponding to the three peptide segments of patient sample 2 mel_38 are as follows:

Table 4

Step 103: Prediction of MHC-I binding neoantigens based on tumor somatic mutations and calculation of related feature values: Based on the tumor somatic mutations and annotation data in step (1), call NetMHCpan, Netchop, OptiType tools to predict MHC-I binding Neoantigens, and calculate the following characteristic values: affinity ranking percentage between mutant peptides and MHC R _m , affinity ranking percentage between unmutated peptides and MHC R _n , peptide clipping and presentation efficiency NC; Take the eigenvalues as an example, the R _m , R _n and NC values calculated corresponding to the three peptides of patient sample 2 mel_38 are as follows:

table 5

Step 104: Extraction of neoantigen-related eigenvalues: for the MHC-I binding neoantigens predicted in step (2), extract all relevant eigenvalues of tumor neoantigens, and the neoantigens and eigenvalues of patient sample 2 mel_38 are shown in Table 4 ,table 5;

Step 105: setting of neoantigen activity scoring function: setting neoantigen activity scoring function for the neoantigen characteristic value extracted in step (3);

Step 106: Neoantigen sorting based on the neoantigen activity scoring function: sort the neoantigens through the neoantigen activity scoring function, and Table 6 shows the scoring (Neo_Score) of the neoantigens verified by the PMHC activity experiment in the patient sample 2 mel_38 and Rank in the corpus

Table 6

Among the three identified neoantigens, FLYNLLTRVY showed the immune activity of activating CD8+ T cells before using tumor vaccines to stimulate DC cells, and the other two had different degrees of immune activity after using vaccines to enhance the immune system. immune activity. We see that QLSCISTYV and FLYNLLTRVY are in the Top 20 (117 total candidate neoantigens). However, KLMNIQQKL, whether in vitro or in the human tumor microenvironment, due to its low expression level, determines its immunogenicity and weak immune response. It ranks 66th in our ranking results, and the experimental results are quite consistent with our predicted ranking. .

In summary, the neoantigen immune activity scoring function we proposed can effectively measure the immune activity of neoantigens, and provide assistance for clinical experiments, tumor research, and immunotherapy.

Claims (4)

1. the prediction of neoantigen immunocompetence and sort method based on tumour neoantigen characteristic value, it is characterised in that including following Step：

(1), the input of WGS/WES, RNA-SEQ sequencing data of tumour-normal sample：Input the full base of tumour-normal sample Because of sequencing data WGS or full sequencing of extron group data WES, transcript profile sequencing data RNA-SEQ；

(2), the prediction of tumour somatic mutation and annotation, the calculating of associated eigenvalue：Sequencing number based on step (1) input According to calling Varscan or Mutect tool analysis calculates tumour somatic mutation, calls VEP (Variant Effect Prediction) instrument complete mutation annotation, call PyClone, Kallisto, Varscan or Mutect instrument calculate as Lower eigenvalue：Mutator clone ratio, mutator expression value TPM, allelic mutation frequency VAF；

(3) prediction of MHC-I combinations neoantigen, the calculating of associated eigenvalue based on tumour somatic mutation：Based in step (2) Tumour somatic mutation and annotation data, call NetMHCpan, Netchop, OptiType instrument prediction MHC-I to combine new Antigen, and calculate such as lower eigenvalue：Peptide fragment is mutated to arrange with MHC affinity sequence percentage, unmutated peptide fragment and MHC affinity Sequence percentage, peptide fragment shearing present efficiency；

(4) extraction of all associated eigenvalues of neoantigen：For the MHC-I combination neoantigens of prediction in step (3), extract swollen All associated eigenvalues of knurl neoantigen；

(5) setting of neoantigen activity scoring functions：For the neoantigen characteristic value of extraction in step (4), setting neoantigen is lived Property scoring functions；

(6) the neoantigen sequence based on neoantigen activity scoring functions：Neoantigen is carried out by neoantigen activity scoring functions Sequence.

2. the method according to claim 11, it is characterized in that：

In step (4), neoantigen associated eigenvalue includes R_m、A、R_n, E, NC, CL, wherein：

R_m- mutation peptide fragment and MHC affinity sequence percentage, are calculated by NetMHCpan；

A-allelic mutation frequency VAF, is calculated by Varscan/Mutect/Strelka2；

R_n- unmutated peptide fragment and MHC affinity sequence percentage, are calculated by NetMHCpan；

E-mutator expression value TPM, is calculated by Kallisto；

NC-peptide fragment shearing presents efficiency, is calculated by netchop；

CL-mutator clone's ratio, is calculated by pyclone.

3. the method according to claim 11, it is characterized in that：

In step (5), the neoantigen Activity Prediction scoring functions of proposition are：

Neo_Score=abundancedissimilarityclonality；

Clonality=NCCL；

Abundance=L (Rm) Atanh (E/k)；

Dissimilarity=(1-L (Rn)/2))；

Wherein：L (x)=1/ (1+e^5(x-2)), tanh (x) is hyperbolic tangent function；

K is transcript gene expression abundance threshold value, default value 1.

4. the method according to claim 11, it is characterized in that：In step (6), newly resisted by neoantigen Activity Prediction function pair It is as follows that original is ranked up algorithmic procedure：

A), for the MHC-I combination neoantigens of all predictions, neoantigen Activity Prediction function Neo_score is called to calculate newly The predicted value of antigen active；

B), the predicted value based on neoantigen activity, is ranked up using quick sorting algorithm to neoantigen；

C), neoantigen ranking results are exported.