Selection of Antibodies / Antibody Fragments

Abstract

The present invention relates to selecting antibodies and / or antibody fragments in a selection method. The subject matter of the present invention is a method, a system and a computer program product for generating a score value for a gene pair encoding the variable region of the light chain and the heavy chain of an antibody and / or antibody fragment. Antibodies and / or antibody fragments can be selected based on the score value.

Description

Technical Field

[0001] The present invention relates to selecting antibodies and / or antibody fragments in a selection method. The subject matter of the present invention is a method, a system and a computer program product for generating a score value for a gene pair encoding the variable region of the light chain and the heavy chain of an antibody and / or antibody fragment. Antibodies and / or antibody fragments can be selected based on the score value. Background Art

[0002] As a defense system against foreign invaders, the human immune system constitutes a complex network of various organs, cell types, and molecules.

[0003] Antibodies (also called immunoglobulins) are proteins from the class of globulins, which are formed in vertebrates as a response to certain substances (the so-called antibodies). Antibodies serve the immune system; they are produced by a class of white blood cells, the B lymphocytes.

[0004] Almost only large molecules or molecules bound to particles (such as lipopolysaccharides on the surface of bacteria) act as antigens. A certain antigen generally induces the formation of only a few specific antibodies, which in most cases recognize only this foreign substance through specific non-covalent bonding.

[0005] The specific binding of antibodies to antigens constitutes an integral part of the defense against invading foreign substances.

[0006] Each antibody is composed of two identical heavy chains (H) and two identical light chains (L), which are linked to each other by covalent disulfide bridges to form a Y-shaped structure. The light chain is composed of a variable region and a constant region. L and C L In contrast, the heavy chain has one variable region and three (IgG, IgA) or four (IgM, IgE) constant regions. H and C H 1. C H 2. C H 3.

[0007] The variable regions of one light chain and one heavy chain constitute the antigen binding sites; these antigen binding sites are therefore of special interest for pharmaceutical, immunological and / or diagnostic purposes. In order to use antibodies and / or antibody fragments, for example in medicine, large libraries of antibodies and / or antibody fragments are created and analyzed.

[0008] A widely used technique for generating and characterizing antibody libraries is the "phage display" method, in which the corresponding protein of interest can be expressed as a fusion polypeptide on a bacterial phage coat protein and selected by binding to a fixed or soluble biotin-labeled ligand (antigen). The phage constructed in this way can be viewed as a compact genetic unit that has combined phenotypic and genotypic properties. The "phage display" technique has been successfully applied to antibodies, antibody fragments, enzymes, DNA-binding proteins, etc.

[0009] In order to use, for example, phage display of antibody libraries, the relevant cells are first isolated from the organism. It involves plasma cells present in particular in the blood, in the bone marrow and in the lymph nodes. From these cells, mRNA is isolated and then transcribed into cDNA.

[0010] The light chain of the antibody (V L ) and heavy chain (V H )'s variable region genes.

[0011] In a special phagemid vector, each genome was linked to a truncated gene of the coat protein pill (minor coat protein) of the M13 phage and used to transform Escherichia coli.

[0012] Thus, the E. coli bacterial expression contains the pilli fusion protein of scFv fragment or Fab antibody fragment.Through the signal peptide, the fusion protein is transported to the periplasm, where the fusion protein is folded into a functional scFv or a Fab fragment connected by a disulfide bridge.The Fv or Fab part first remains anchored in the E. coli inner membrane through the pilli fragment and is attached to the capsid when the phage is assembled.

[0013] Through the coat protein pill, which is usually responsible for bacterial infection, after co-infection with the M13 helper phage, the functional antibody fragments are loaded into the coat of the newly formed phage during its maturation process. At the same time, the phagemid containing the relevant gene information of the corresponding antibody fragment is loaded into the interior of the newly formed phage. That is, each of these recombinant phages theoretically has different antibody fragments on its surface and at the same time has the relevant genes (V L and V H ), similar to the billions of B cells in the body.

[0014] In so-called biopanning, "binding" phages can be selected from a billion-fold background of irrelevant phages through interaction with an immobilized ligand (antigen) via antibody fragments exposed on the surface.

[0015] Bio-panning generally involves multiple selection cycles (panning rounds). This generally involves exposing a substrate to a phage display library so that the binding of several phages can be performed. Non-specifically bound and weakly bound phages are washed away. Phages that are still bound and therefore specific after cleaning are subsequently separated (eluted). The eluted phages are propagated and re-exposed to the substrate in further panning rounds until an enriched population of effectively bound phages is obtained.

[0016] At the end of the selection process, the relevant antibody genes can be simply isolated from the isolated phage and sequenced. At this time, sequencing obtains information about the spectrum of the antibody (fragment). Other methods for selecting antibodies and / or antibody fragments are described in the literature (see, for example, Sai T. Reddy et al.: Monoc / onal antibodies isolatedwithout screening by anaylzing the variable-gene repertoire of plasma cells, Nature Biotechnology Vol.28No.9, Sept.2010, 965-971).

[0017] The standard method for determining sequences is the dideoxy method (chain termination synthesis) according to Sanger. The length of the DNA segments (reads) sequenced in this way can reach more than 1000 base pairs, but this is associated with high costs and time consumption. An alternative is next-generation sequencing (NGS). Due to the detection of many parallel and spatially separated sequencing reactions on a very small area, significantly higher throughput is quickly achieved with shorter read lengths compared to standard methods. NGS has achieved 10 6 This means that it is possible to determine the diversity and quality of an entire library.

[0018] In the biopanning selection method, the final pool is enriched for specific binding antibody fragments, but it always contains a large number of different antibody fragments. This number is too large to study all antibody fragments in detail; selection must be performed. In many cases, the V value of the antibody fragments in the final pool is used to select the fragments that are most likely to be present. H The number of genes sorts the antibody fragments. Generally speaking, the more antibody fragments that have been enriched, the higher their V H The greater the number of genes in the final pool.

[0019] However, in the last pool, V H The number of genes is not a sufficient basis for selecting antibodies of particular interest and / or value.

[0020] It would therefore be desirable to provide better aids in identifying promising antibody fragments. Summary of the invention

[0021] This object is achieved by the subject matter of the independent claims. Preferred embodiments are found in the dependent claims, the description and the drawings.

[0022] In a first aspect, the present invention provides a method comprising the following steps:

[0023] -Receive involves V L Gene and V H Gene features, where V L Gene and V H The genes encode the variable regions of the light chain and the heavy chain of antibodies and / or antibody fragments, wherein the antibodies and / or antibody fragments are derived from a selection method, wherein the selection method comprises a plurality of selection cycles followed one after another, wherein the result of each selection cycle is a pool of antibodies and / or antibody fragments,

[0024] - For these pools V L Gene and V H Gene pairs are used to construct feature vectors, where each pair of V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features:

[0025] ·V L The frequency of occurrence of a gene in the corresponding pool,

[0026] ·V H The frequency of occurrence of a gene in the corresponding pool,

[0027] ·V L The frequency of the gene in the previous pool,

[0028] ·V H The frequency of the gene in the previous pool,

[0029] ·V H The frequency of occurrence of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools,

[0030] ·V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool,

[0031] - Identify the enrichment patterns in these feature vectors of these pools, where the enrichment pattern represents the enrichment patterns in the selection method using the corresponding VL Gene and V H The strength of gene enrichment for the antibody and / or antibody fragment,

[0032] - Based on this enrichment pattern, V L Gene and V H The scores for the gene pairs are related to the corresponding enrichment strengths,

[0033] -Optionally: Get V based on these score values L Gene and V H A ranked list of gene pairs,

[0034] - Output these score values ​​and / or ranked lists to the user.

[0035] The present invention further provides a system, the system comprising:

[0036] Input unit,

[0037] Feature vector generation unit,

[0038] Enrichment analysis unit,

[0039] Fractional value calculation unit,

[0040] A sorting unit, and

[0041] Output unit,

[0042] - wherein the input unit is configured to obtain a voltage related to V L Gene and V H Gene characteristics, V L Gene and V H The genes encode the variable regions of the light chain and the heavy chain of antibodies and / or antibody fragments, wherein the antibodies and / or antibody fragments are derived from a selection method, wherein the selection method comprises a plurality of selection cycles followed one after another, wherein the result of each selection cycle is a pool of antibodies and / or antibody fragments,

[0043] wherein the feature vector generating unit is configured to generate V in the pool based on the obtained features L Gene and V H The feature vector of gene pairs, where V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features:

[0044] ·V L The frequency of occurrence of a gene in the corresponding pool,

[0045] ·V HThe frequency of occurrence of a gene in the corresponding pool,

[0046] ·V L The frequency of the gene in the previous pool,

[0047] ·V H The frequency of the gene in the previous pool,

[0048] ·V H The frequency of occurrence of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools,

[0049] ·V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool,

[0050] - wherein the enrichment analysis unit is configured to identify an enrichment pattern in the feature vector of the pool, wherein the enrichment pattern represents the enrichment pattern of the pool in the selection method using the corresponding V L Gene and V H The strength of gene enrichment for the antibody and / or antibody fragment,

[0051] - wherein the score value calculation unit is configured to calculate V based on the enrichment pattern L Gene and V H The scores for the gene pairs are related to the corresponding enrichment strengths,

[0052] - wherein the ranking unit is preferably configured to calculate V based on the score value L Gene and V H A ranked list of gene pairs,

[0053] - wherein the output unit is configured to output the score values ​​and / or the ranked lists to a user.

[0054] The invention further provides a computer program product comprising a data carrier and a program code which is stored on the data carrier and which, when loaded into a working memory of a computer system, causes the computer system to perform the following steps:

[0055] -Receive involves V L Gene and V H Gene features, where V L Gene and V H The genes encode the variable regions of the light chain and the heavy chain of antibodies and / or antibody fragments, wherein the antibodies and / or antibody fragments are derived from a selection method, wherein the selection method comprises a plurality of selection cycles followed one after another, wherein the result of each selection cycle is a pool of antibodies and / or antibody fragments,

[0056] - For these pools V L Gene and V H Gene pairs are used to construct feature vectors, where each pair of V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features:

[0057] ·V L The frequency of occurrence of a gene in the corresponding pool,

[0058] ·V H The frequency of occurrence of a gene in the corresponding pool,

[0059] ·V L The frequency of the gene in the previous pool,

[0060] ·V H The frequency of the gene in the previous pool,

[0061] ·V H The frequency of occurrence of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools,

[0062] ·V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool,

[0063] - Identify the enrichment patterns in these feature vectors of these pools, where the enrichment pattern represents the enrichment patterns in the selection method using the corresponding V L Gene and V H The strength of gene enrichment for the antibody and / or antibody fragment,

[0064] - Based on this enrichment pattern, V L Gene and V H The scores for the gene pairs are related to the corresponding enrichment strengths,

[0065] -Optionally: Get V based on these score values L Gene and V H A ranked list of gene pairs,

[0066] - Output these score values ​​and / or ranked lists to the user.

[0067] The present invention will be explained in detail below without distinguishing the inventive subject matter. The following explanation should be applied to all the inventive subject matters in a similar manner, regardless of the context.

[0068] When multiple steps are mentioned in a certain order in the present description or claims, this does not necessarily mean that the invention is limited to the mentioned order. Rather, it is also conceivable that the steps can also be performed in a different order or in parallel with each other; unless one step is based on another step, it is necessarily required that this step is performed after the underlying step (but this will become clear in the specific case). The mentioned order therefore forms a preferred embodiment.

[0069] The starting point of the present invention is a library of antibodies and / or antibody fragments. These antibodies and / or antibody fragments are sent to a selection method in order to select antibodies and / or antibody fragments based on their phenotypic properties. The selection method can be, for example, a bio-panning method.

[0070] The selection method in the sense of the present invention comprises a plurality of selection cycles in succession. The result of each selection cycle is a pool with antibodies and / or antibody fragments. The purpose of the selection method is to enrich for antibodies and / or antibody fragments that specifically bind. This means that any pool in a selection cycle has a higher concentration of antibodies and / or antibody fragments that specifically bind than the previous pool.

[0071] Figure 1 The example of the selection hierarchy with three layers (I, II, III) is shown. In the first layer (I), the library of antibodies and / or antibody fragments is sent to the first selection cycle. The result of the first selection cycle is the pool of the selected antibodies and / or antibody fragments. In the second layer (II), the selected antibodies and / or antibody fragments are sent to the second selection cycle (a and b). The rotation cycle in the second layer is generally different selection cycles. The result of the selection in the second layer is then two pools of the selected antibodies and / or antibody fragments. The antibody and / or antibody fragment pool obtained from the right branch (IIb) of the second layer is subjected to two selection cycles (a and b) again in the third layer (III), which produces two pools again. When completely passing through the selection hierarchy (e.g., from the first layer through the second layer to the third layer) from top to bottom, antibodies and / or antibody fragments are generally subjected to increasing selection pressure. If completely passing through the selection hierarchy (I→IIb→IIIb) from top to bottom, the diversity of antibodies and / or antibody fragments in the corresponding pools is reduced.

[0072] It is conceivable that different targets are used in the selection cycle. Therefore, it is common, for example, to study the binding of antibodies not only with respect to human targets but also with respect to murine targets when searching for pharmaceutically effective antibodies or when searching for antibodies for diagnostic purposes. Correspondingly, the selection method usually has at least two branches, wherein the antibody is subjected to increasing selection pressure for a first target (e.g., a murine target) in one branch, and is subjected to increasing selection pressure for a second target (e.g., a human target) in a second branch (see, e.g., WO0220822A2).

[0073] The selection method preferably comprises at least two, preferably at least 3, still more preferably at least 4, most preferably at least 5 selection layers.

[0074] After a selection cycle, the genes of the selected antibodies and / or antibody fragments are sequenced in the usual manner to identify them.

[0075] Sequencing involves determining the V sequences encoding the variable regions of the light chains of antibodies and / or antibody fragments. L Genes and V genes encoding the variable regions of the heavy chains of antibodies and / or antibody fragments H Genes. At the same time, their occurrence frequencies (counts) in the corresponding pools were determined.

[0076] It is conceivable that such a sequencing step is performed after each selection cycle. It is also conceivable that such a sequencing step is performed only after a hierarchy of multiple cycles has been completely passed. It is conceivable that a certain degree of enrichment of specifically bound antibodies and / or antibody fragments should first be achieved and then characterized by sequencing.

[0077] In another step, the V L and V H Gene characterization and generation of V L Gene and V H The feature vector of gene pairs, where V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment.

[0078] In general, a feature vector (English: feature vector) combines the parameterizable properties (features) of an object (preferably in numerical form) in a vector manner. The various features that are specific to the object constitute the different dimensions of this vector. The entirety of possible feature vectors is called the feature space. Feature vectors make automatic classification easier, for example, because they significantly reduce the features to be classified.

[0079] In this case, the target is a V in a pool. L Gene and V H Gene pair, where V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment.

[0080] A method for identifying V sequences encoding variable regions belonging to the same antibody and / or antibody fragment L Gene and V HThe method of gene pairing is described in the international patent application with publication number WO 2019206729, the entire contents of which are incorporated into this specification by reference. Because the above patent application has not been published on the priority date of the protection right of the present invention, it is not necessary to identify V genes encoding variable regions belonging to the same antibody and / or antibody fragment. L Gene and V H Basic aspects of gene pairs are described in detail at the end of this specification.

[0081] The feature vector represents V in a pool L Gene and V H Gene pairs. For the same pair in another pool (e.g., the previous pool), a separate feature vector is generated. Similarly, for another pair in the same pool, a separate feature vector is also generated. According to the present invention, in the feature vector for a pair in a pool, not only information about the pair in the pool is entered, but also, for example, information about the same pair in the previous pool is entered.

[0082] One V for one pool L Gene and V H The feature vector of a gene pair includes at least the following features:

[0083] ·V L The frequency of occurrence of the gene in the corresponding pool (preferably in V L The absolute number of genes in the pool (in VL form or in normalized form),

[0084] ·V H The frequency of occurrence of the gene in the corresponding pool (preferably in V H The absolute number of genes in the pool (VH format or in normalized format),

[0085] ·V L The frequency of occurrence of the gene in the previous pool (preferably in V L The absolute number of genes in the previous pool (prevVL) or in normalized form)

[0086] ·V H The frequency of occurrence of the gene in the previous pool (preferably in V H The absolute number of genes in the previous pool (in prevVH form or in normalized form),

[0087] ·V H The frequency of occurrence of genes in the corresponding pool is related to V L The absolute difference (diff) between the occurrence frequencies of genes in the corresponding pools (diff = |VH-VL|),

[0088] ·V HThe frequency of gene occurrence in the previous pool is related to V L The absolute difference between the occurrence frequencies of genes in the previous pool (prevDiff) (prevDiff = |prevVH-prevVL|),

[0089] In a pool involving V L Gene and V H An important characteristic of a gene is its frequency of occurrence in the corresponding pool (preferably in the form of absolute numbers VH and VL or in normalized form). Preferably, VH and VL are sequenced by next generation sequencing with the aid of paired recognition sequences (primer pairs). H and V L The fragmented genetic information of the sequence is assembled to form a complete V H and V L The frequency of occurrence is obtained by sifting the strand (or part thereof). Unique sequences are then counted and output in the form of DNA counts.

[0090] The normalization of the occurrence frequency can be performed based on the following formula, for example:

[0091] rpmVH=(VH / sumVH)*1,000,000

[0092] rpmVL=(VL / sumVL)*1,000,000

[0093] Where rpmVH is V H The normalized frequency of occurrence of a gene in the observed pool,

[0094] Where rpmVL is V L The normalized frequency of occurrence of a gene in the observed pool,

[0095] Where sumVH is the total number of V in the observed pool. H The sum of the absolute frequency of occurrence of genes VH,

[0096] Where sumVL is the total number of V in the observed pool. L The sum of the absolute occurrence frequencies VL of genes.

[0097] Other important features are V L The frequency of the gene in the previous pool and V H The frequency of occurrence of a gene in the previous pool. The previous pool is preferably understood to mean a pool that is located above the pool under consideration in the selection hierarchy (that is, directly before the pool under consideration in the order of the selection cycles passed). The antibodies / antibody fragments isolated from the previous pool are fed to the selection cycle, the result of which is the pool under consideration. The V in the previous pool L Gene and V HThe frequency of occurrence of genes can also be used in the feature vector in absolute or normalized form.

[0098] Other important features are V H The frequency of occurrence of a gene in the observed pool is related to V L The absolute difference (diff) between the occurrence frequencies of genes in the observed pool (diff = |VH-VL|), and V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference (prevDiff) between the occurrence frequencies of genes in the previous pool (prevDiff = |prevVH - prevVL|).

[0099] In addition, other features can be incorporated into the feature vector, for example:

[0100] · In the observed pools, different V H Absolute number of genes (frequency of occurrence): numVH

[0101] In the observed pools, different V L Absolute number of genes (occurrence frequency): numVL

[0102] The V that appears most frequently in the observed pool H Absolute number of genes (frequency of occurrence): maxVH

[0103] The V that appears most frequently in the observed pool L Absolute number of genes (occurrence frequency): maxVL

[0104] In the observed pool V H The relative frequency of the gene (relative to the most frequent V gene in the observed pool) H Number of genes): relVH = VH / maxVH

[0105] In the observed pool V L The relative frequency of the gene (relative to the most frequent V gene in the observed pool) L Number of genes): relVL = VL / maxVL

[0106] In the observed pool V H The frequency of occurrence of a gene is related to the V LThe relative difference (relative distance) between the occurrence frequencies of genes (relative to the occurrence frequency of the gene with higher occurrence frequency in the observed pool): reldiff = |VH-VL| / max(VH, VL), where for VH>VL, max(VH, VL) = VH and for VL≥VH, max(VH, VL) = VL,

[0107] The number of selection cycles (layers) that have been completed before the observed selection cycle: prevnum

[0108] ·V H Normalized frequency of occurrence of a gene in the observed pool: rpmVH = (VH / sumVH)*1000000, where sumVH is the sum of all VH in the observed pool. H The sum of the gene occurrence frequencies VH,

[0109] ·V L Normalized frequency of occurrence of a gene in the observed pool: rpmVL = (VL / sumVL)*1000000, where sumVL is the sum of all VL in the observed pool. L The sum of the gene occurrence frequencies VL,

[0110] ·V H Normalized frequency of occurrence of the gene in the previous pool: prevRpmVH = (prevVH / prevSumVH) * 1000000, where prevSumVH is the normalized frequency of occurrence of the gene in the previous pool. H The sum of the gene occurrence frequencies VH,

[0111] ·V L Normalized frequency of occurrence of genes in the previous pool: prevRpmVL = (prevVL / prevSumVL) * 1000000, where prevSumVL is the normalized frequency of occurrence of all V genes in the previous pool. L The sum of the gene occurrence frequencies VL,

[0112] V from the first selection cycle to the second selection cycle H Relative change in gene number: prevRelDiffV H =(|VH-prevVH|) / max(VH,prevVH), where for VH>prevVH, max(VH,prevVH)=VH and for prevVH≥VH, max(VH,prevVH)=prevVH,

[0113] V from the first selection cycle to the second selection cycle L Relative change in gene number: prevRelDiffV L=(|VL-prevVL|) / max(VL,prevVL), where for VL>prevVL, max(VL,prevVL)=VL and for prevVL≥VL, max(VL,prevVL)=prevVL.

[0114] ·The observed pool V H The normalized frequency of genes is the same as the V in the previous pool. H The logarithm of the ratio of the normalized frequency of occurrence of genes with base 2: logRpmVH = log2(rpmVH / prevRpmsVH)

[0115] ·The observed pool V L The normalized frequency of genes is the same as the V in the previous pool. L The base 2 logarithm of the ratio of the normalized frequency of occurrence of genes: logRpmVL=log2(rpmVL / prevRpmsVL).

[0116] It is conceivable that in addition to the mentioned features, other information is obtained / determined and input into the generation of the feature vector. For example, V from a pool of other selection cycles can be obtained. H Gene and V L The frequency of occurrence of the gene is used as a feature and is input into the generation of the feature vector. In addition, the features of the parameters of the selection level and / or the selection cycle can be obtained and are input into the generation of the feature vector. For example, the matrix used in the selection cycle, the parameters (concentration, temperature, medium, etc.) for performing the sequencing cycle, the structure of the sequencing level (the number of layers and / or branches), etc. are also included.

[0117] The eigenvector usually includes the corresponding V L The unique identifier of the gene and the corresponding V H The unique identifier of the gene. The unique identifier is used to identify the observed V H Genes and observed V L The unique identifier can be, for example, a gene sequence, name, identification number, letter identification code, or a unique label V L Gene or V H The unique identifier is therefore mainly used to process the results and assign the results to the corresponding V H Gene and V L It is also conceivable that instead of V belonging to a pair L Gene and V belonging to the pair H The genes each individually provide a unique identifier, but the pair is characterized by one unique identifier.

[0118] V with corresponding identifiers x and y L Gene and V H Examples of preferred feature vectors for gene pairs are:

[0119] <x, y, VL, VH, diff, relDiff, prevVL, prevVH, prevDiff>

[0120] V with corresponding identifiers x and y L Gene and V H Examples of particularly preferred feature vectors for gene pairs are:

[0121] <x, y, rpmVL, rpmVH, diff, relDiff, prevDiff, prevRpmVH, prevRpmVL, prevNum, logRpmVH, logRpmVL>

[0122] Naturally, the order in which the features appear in the feature vectors described herein may also be different from that shown.

[0123] In the next step, the enrichment pattern is identified in the data set constructed by the feature vector. That is, unlike what is common in the prior art, the enrichment of antibodies or antibody fragments is determined not only based on the frequency of occurrence of genes in the last pool, but also based on the features from the last pool and at least one previous pool. Feature vectors from at least three, still more preferably from at least four pools that are connected to each other are preferably used to identify the frequency of occurrence pattern.

[0124] In each case, for V L Gene and V H Gene pairs identified enrichment patterns, where the V L Gene and V H The genes encode the variable regions belonging to the same antibody and / or antibody fragment. The enrichment pattern indicates the enrichment of the corresponding V L Gene and V H The gene is enriched for the strength of the antibody and / or antibody fragment. That is, the enrichment pattern indicates the binding strength of the corresponding antibody / antibody fragment to the antigen (or multiple antigens used in the selection method) used in the selection method: the higher the enrichment, the higher the binding strength.

[0125] Preferred Identification V L Gene and V H The gene pairs have enrichment patterns against multiple different targets, preferably against one or more murine targets and one or more human targets.

[0126] Then, based on the enrichment pattern, each V LGene and V H Gene pairs are calculated with scores that quantitatively reflect the enrichment of the corresponding antibody / antibody fragment. Preferably separate scores are calculated for enrichment against different targets (e.g. one score is calculated for enrichment against a murine target and one score is calculated for enrichment against a human target).

[0127] The approach used to identify enrichment patterns is to track V from one pool of the selection method to the next. L Gene and V H However, according to the present invention, not only V L Gene and V H The frequency of occurrence of genes, and also additionally observe one or more of the above-mentioned other characteristics. In addition, not only the last pool is observed, but also at least two, preferably at least three, still more preferably at least four pools that are connected to each other are observed.

[0128] In order to identify the enrichment pattern, preferably multivariate data analysis methods are used.

[0129] To this end, dimensionality reduction is preferably performed in the first step. It is preferred to determine V in a pool based on the feature vector. L Gene and V H Factors of gene pairs, which describe the characteristics of these pairs in the pool with fewer variables than eigenvectors. Commonly used methods are, for example, principal component analysis (English: Principal Component Analysis (PCA)). However, other methods of multivariate data analysis are also conceivable.

[0130] In principal component analysis, a large number of variables are approximated by a relatively small number of linear combinations ("principal components") that are as meaningful as possible. The aim of principal component analysis is to project the data points in a p-dimensional space into a q-dimensional subspace in such a way that as little information as possible is lost and the currently existing redundancies are simplified in the form of correlations in the data points. In this case, the positive integers p and q give the dimensions of the respective spaces. The principal axis transformation is carried out mathematically: by transforming into a vector space with a new basis, the correlation of the multidimensional features is minimized. The principal axis transformation can be given by an orthogonal matrix consisting of the eigenvectors of the covariance matrix. More details can be found in the extensive specialist literature on this subject (see, for example, H.-F. Eckev et al.: Multivariate Statistik, Gabler Verlag 2002, ISBN-13: 978-3-409-11969-6).

[0131] The variation of the characteristic vectors is preferably approximated by a number of factors which cover at least 70% of the variability of the original data.

[0132] In a preferred embodiment, 5 to 7 variables of the eigenvector are approximated by 2 or 3 factors (eg principal components).

[0133] Based on these factors, in the next step, V L Gene and V H Gene pairs are scored. Score values ​​are preferably generated over all pools (with specific targets, such as murine targets or human targets) for which there are feature vectors for the corresponding pairs. For each pair, this is at least two, preferably at least 3, and still more preferably at least 4 feature vectors (per target).

[0134] The score value and V caused by the selection method L Gene and V H The score is associated with the enrichment of the gene (or the corresponding antibody / antibody fragment). Generally, the higher the score, the more obvious the enrichment (positive correlation).

[0135] The score reflects the V L Gene and V H Evolution of the factors of a gene pair from the first (observed) pool to the last (observed) pool of the selection method.

[0136] In a preferred embodiment, the sum of the following vectors is calculated: these vectors are generated when the coordinates of the factors (principal components) along the pools are advanced from the first pool to the last pool. A vector is generated for each step from the factor coordinates of a pool to the factor coordinates of the next pool. The sum of all vectors is a sum vector, which gives a point of these factors in the coordinate system starting from the origin of the coordinate system. For example, the length of this vector can be given as a fractional value. It is also conceivable that the fractional value is calculated by the product of the coordinates of the point pointed to by the sum vector in the coordinate system of these factors. It is conceivable that these coordinates are weighted before multiplication. For example, it is conceivable that the first factor (first principal component) should be given a higher weight than the second factor (second principal component). Correspondingly, the coordinates of the first factor can be multiplied with a higher weight than the coordinates of the second factor. In a preferred embodiment, the coordinates of the individual factors are weighted identically.

[0137] In the next step, the score values ​​are arranged in a ranked list. To this end, the score values ​​are usually ranked corresponding to their size, for example from the largest to the smallest score value or vice versa. If the score value is positively correlated with the enrichment, a large score value indicates that the V of those antibodies / antibody fragments with particularly significant specific enrichment H -V L Yes. These antibodies / antibody fragments are therefore of particular interest for pharmaceutical, immunological and / or diagnostic purposes. The ranked list is output to the user, for example, by means of a display screen.

[0138] As already explained, different score values ​​can be determined for antibodies and / or antibody fragments, for example one score value is calculated for enrichment against a murine target and one score value is calculated for enrichment against a human target. A ranked list can be generated for each of these different score values.

[0139] The calculated score values ​​and / or the ranked list can be output to a user, for example, by means of a display screen.

[0140] The steps of the method of the present invention can be implemented with the aid of one computer system or multiple computer systems.

[0141] A "computer system" is a system for electronic data processing, which processes data with the aid of programmable calculation rules. Such a system usually includes a "computer", that is, a unit including a processor for performing logical operations, and peripheral devices.

[0142] In computer technology, the term "peripheral device" refers to all devices that are connected to a computer and are used to control the computer and / or serve as input and output devices. Examples are monitors (displays), printers, scanners, mice, keyboards, drives, cameras, microphones, speakers, etc. In computer technology, internal interfaces and expansion cards are also considered peripheral devices.

[0143] Computer systems today are often divided into desktop PCs, portable PCs, laptops, notebooks, netbooks and tablet PCs as well as so-called handheld devices (eg smartphones); all of these devices can be used to implement the present invention.

[0144] Input to a computer may be via input devices such as a keyboard, mouse, microphone, network connection, external data storage, and / or the like. Output is typically via a display screen (monitor), on a printer, via speakers, and / or by storage on a data storage device.

[0145] The system for performing the present invention is configured to implement a series of operations. In this specification, these operations are assigned to specific units: input unit, feature vector generation unit, enrichment analysis unit, score value calculation unit, sorting unit and output unit. However, it is conceivable that these operations are implemented by a (single) central processing unit (English: Central Processing Unit (CPU)) of a computer system. It is also conceivable that these operations are distributed to different computers (computer systems), where one computer (computer system) can implement one or more operations of the unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0146] The invention is explained in more detail below with the aid of the drawings, without intending to restrict the invention to the features and feature combinations shown in the drawings.

[0147] In the attached picture:

[0148] Figure 1 An example of a selection hierarchy with three levels (I, II, III) is shown. Figure 1 Described in detail above.

[0149] Figure 2 The biopanning cycle is schematically illustrated taking one selection cycle as an example.

[0150] In a first step a library (1) of bacteriophages (1a, 1b, 1c) is provided, in which antibodies and / or antibody fragments (10, 11, 12) are expressed as fusion polypeptides on a phage coat protein.

[0151] In addition, a matrix (2) with immobilized antigens and / or antigen fragments is provided. The immobilized antigens and / or antigen fragments have binding sites (20) to which antibodies and / or antibody fragments (10, 11, 12) of phages (1a, 1b, 1c) can bind. The antigens / antigen fragments can be mouse and / or human antigens / antigen fragments.

[0152] In step A of the biopanning cycle, phage are incubated with the substrate.

[0153] During this time, an interaction occurs between the antibody / antibody fragment and the antigen. In the present case, the antibody / antibody fragment (1c) exactly matches the binding site (20) of the immobilized antigen; the interaction and the resulting binding are relatively strong (stronger than in the case of other antibodies / antibody fragments).

[0154] Phage that do not bind or only weakly bind to the matrix are separated (washed away) in step B of the biopanning cycle. Phage that bind more strongly are retained.

[0155] The more strongly bound phage are released from the matrix in step C of the biopanning cycle.

[0156] Phage released from the matrix are propagated in step D of the biopanning cycle. The result is a new phage library that can be re-exposed to the matrix.

[0157] A portion of the phage released from the matrix can be sequenced for the genes of the antibody and / or antibody fragment (step E). This involves determining the V sequences encoding the variable regions of the light chains of the antibodies and / or antibody fragments. L Genes and V genes encoding the variable regions of the heavy chains of antibodies and / or antibody fragments H Genes and their frequency of occurrence.

[0158] Figure 3 The biopanning procedure with three cycles (I, II, III) is schematically shown. If, as in the present case, the same matrix is ​​used each time, the strongest binding antibodies / antibody fragments are further enriched with each cycle. Step A I , B I , C L , D I 、E I or A II , B II , C II , D II 、E II or A III , B III , C III , D III 、E III Corresponds to Figure 2 Follow steps A, B, C, D, and E in the code.

[0159] After each cycle (I, II, III), V H Gene and V L Genes and their frequencies (step E I 、E II and E III ).

[0160] Figure 4 Schematically showing Figure 3 The same biopanning method as in the above, with cycles I, II and III and step A I , A II , A III , B I , B II , B III , C I , C II , C III , D I , D II , D III 、E I 、E II and E III In each case, step E I 、E II and E LLL The results are respectively V L Gene Pool {V L 1 , V L 2 , …, V L n} and a V H Gene Pool {VH 1 , V H 11 , …, V H m}; Each gene can be identified based on a unique identifier (1, 2, ... to n or i, ii, ... to m).

[0161] Figure 5 An embodiment of the system of the present invention is schematically shown. The system (100) includes an input unit (110), a feature vector generation unit (120), an enrichment analysis unit (130), a score value calculation unit (140), a sorting unit (150) and an output unit (160).

[0162] Through the input unit (110), the V L Gene and V H The features of genes encoding the variable regions of the light and heavy chains of the antibodies and / or antibody fragments from the selection method are input into the system (100). The features M are passed to the feature vector generation unit (120). The feature vector generation unit (120) generates the V in the corresponding pool of the selection method based on the features M. L Gene and V H The feature vector MV of the gene pair, where V L Gene and V H The genes encode the variable regions belonging to the same antibody and / or antibody fragment. The feature vector MV is passed to the enrichment analysis unit (130). The enrichment analysis unit (130) generates factors F based on the feature vector MV, which describe these V genes in the pool with fewer variables than the feature vector. L -V H The factor F is passed to the score value calculation unit (140). The score value calculation unit (140) calculates V based on the factor F. L -V H The score value S is passed to the sorting unit (150). The sorting unit (150) generates V based on the score value S. L -V H The sorted list R is transmitted to the output unit (160). The output unit (160) outputs the sorted list R to the user P.

[0163] The system (100) of the present invention may be designed as a computer system (eg, a desktop computer, a tablet computer, a smart phone, a server) or a group of computer systems, for example.

[0164] Figure 6Another embodiment of the system of the present invention is schematically shown. The system (100) comprises an input unit (110), a control and calculation unit (170) and an output unit (160).

[0165] The control and computing unit (170) is configured to:

[0166] - causing the receiving unit (110) to receive the V L Gene and V H The gene feature M, where V L Gene and V H The genes encode the variable regions of the light and heavy chains of antibodies and / or antibody fragments, wherein the antibodies and / or antibody fragments are derived from a selection method, wherein the selection method comprises a plurality of selection cycles followed one after another, wherein the result of each selection cycle is a pool of antibodies and / or antibody fragments.

[0167] - For these pools V L Gene and V H Gene pairs generate feature vectors, where V L Gene and V H The genes encode the variable regions belonging to the same antibody and / or antibody fragment.

[0168] - Based on these feature vectors, generate these V in a pool L Gene and V H factors of gene pairs, where these factors describe the characteristics of these pairs in a pool with fewer variables than eigenvectors,

[0169] - Based on these factors, calculate V L Gene and V H The score of the gene pair,

[0170] -Generate V based on these score values L Gene and V H The sorted list R of gene pairs,

[0171] - Cause the output unit (160) to output the sorted list R to the user.

[0172] The system (100) of the present invention may be designed as a computer system (eg, a desktop computer, a tablet computer, a smart phone, a server) or a group of computer systems, for example.

[0173] Figure 7 A preferred embodiment of the method of the present invention is shown in the form of a flow chart. The method (200) comprises the following steps:

[0174] (210) Receiving V L Gene and V HThe gene feature M, where V L Gene and V H The genes encode the variable regions of the light chain and the heavy chain of antibodies and / or antibody fragments, wherein the antibodies and / or antibody fragments are derived from a selection method, wherein the selection method comprises a plurality of selection cycles followed one after another, wherein the result of each selection cycle is a pool of antibodies and / or antibody fragments,

[0175] (220) Based on these features M, for V in these pools L Gene and V H Gene pairs are used to construct feature vectors MV, where each pair of V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features:

[0176] ·V L The frequency of occurrence of a gene in the corresponding pool,

[0177] ·V H The frequency of occurrence of a gene in the corresponding pool,

[0178] ·V L The frequency of the gene in the previous pool,

[0179] ·V H The frequency of the gene in the previous pool,

[0180] ·V H The frequency of occurrence of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools,

[0181] ·V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool,

[0182] (230) Based on these feature vectors MV, these V L Gene and V H Factors F of gene pairs, where these factors F describe the characteristics of these pairs in a pool with fewer variables than the eigenvectors,

[0183] (240) Based on these factors F, V L Gene and V H The score value S of the gene pair,

[0184] (250) Based on these fractional values ​​S, we obtain V L Gene and V H The sorted list R of gene pairs,

[0185] (260) Output these sorted lists R to user P.

[0186] Figure 8 The results of the principal component analysis of the eigenvector are shown in the form of an image. F1 is the first principal component and F2 is the second principal component. The image shows the proportion of the variables of the eigenvector in the principal components F1 and F2. For example, it can be seen that V H The number of genes VH has a larger proportion in principal component F2, while V L The number of genes VL has a larger proportion in principal component F1. With the help of principal component F1, 40.9% of the variability of the data can be explained; with the help of principal component F2, 21.7% of the data can be explained.

[0187] Fig. 9 Schematically showing a V L -V H The value of the factor for θ along the selection method evolves. The selection method includes six selection cycles (I, II, III, IV, V, VI) that are directly followed by each other. That is, the selection method starts with selection cycle I and ends with selection cycle VI. For the observed V L -V H Yes, at the end of a selection cycle, feature vectors are generated for each pool and the factors are generated based on these feature vectors. Fig. 9 The variables of factors F1 and F2 for these individual pools are shown.

[0188] Fig.10 Schematically showing a V L -V H The generation of the score value of Fig. 9 Since the selection method runs along the selection cycle I → II → III → IV → V → VI, the corresponding vectors are marked in the coordinate system of factors F1 and F2. The sum of the vectors results in the dashed sum vector SV. For example, the length of the sum vector SV can be calculated as V L -V H Possible fractional values ​​for. It is also conceivable that, when the starting point of the vector is placed at the origin of the coordinate system, the product f1·f2 of the coordinates (f, f2) of the point indicated by the sum vector SV is calculated as the fractional value. It is conceivable that the addends in the sum vector are weighted before they are added. By such weighting, for example, selection cycles that result in a greater enrichment of antibodies and / or antibody fragments can be weighted more than selection cycles that result in less enrichment.

[0189] The fractional value is used to quantify the movement of the variable of the factor from the origin of the coordinate system along the selection hierarchy in the direction of higher values ​​of one or more factors. The described (weighted) sum vector quantifies, for example, the movement of points I to VI in the direction of higher values ​​for F1 and F2. Other fractional values ​​that quantify such movement are also conceivable.

[0190] Fig.11 Example shows V L -V H Comparison of the score value with the binding strength of the corresponding antibody / antibody fragment that has been determined in an enzyme-linked immunosorbent assay (ELISA). The binding strength is plotted on the ordinate and the score value on the abscissa. These values ​​are positively correlated to each other.

[0191] Fig.12 The results show that the method of the present invention can identify promising antibodies compared with conventional methods (based on V H The frequency of occurrence of CDR3 was selected (based on V H The frequency of occurrence of genes is selected for comparison. Fig.12 The three performance values ​​are graphically shown in Figure 1; A: Sensitivity (SN), B: Specificity (SP), C: Area under the so-called receiver operating characteristic curve (AUC). The closer these values ​​are to the value 1, the better the prediction result.

[0192] The entry "Random" plots a random antibody selection. Here, its value is naturally at 0.5.

[0193] The entry "Laboratory Standard" shows the standard according to the conventional method (based only on V H The results of selecting candidates based on the frequency of occurrence of CDR3 counts).

[0194] The remaining entries indicate the results of predicting promising candidates for different feature vectors according to the present invention. All predictions are better than conventional methods. DETAILED DESCRIPTION

[0195] V L -V H Identification of the right

[0196] The feature vector generated when implementing the present invention is for V in a pool L Gene and V H Here, V in a pair L Gene and V H The gene pairs encode the variable regions of the light and heavy chains of the same antibody and / or antibody fragment. These pairs are also referred to as V L -V H An example of how these pairs can be identified is described below.

[0197] The starting point is a selection method that has already been performed on a library of antibodies / antibody fragments. The library of antibodies and / or antibody fragments is fed into a selection method in order to select antibodies and / or antibody fragments based on their phenotypic properties. The selection method can be, for example, a biopanning method. The selection method comprises at least two selection cycles that are followed by one another, i.e. a first selection cycle and a second selection cycle. After a selection cycle, the genes of the selected antibodies and / or antibody fragments are sequenced in the usual manner. The purpose of the sequencing is to determine the V sequences encoding the variable regions of the light chains of the antibodies and / or antibody fragments. L Genes and V genes encoding the variable regions of the heavy chains of antibodies and / or antibody fragments H Gene.

[0198] In another step of the method of the invention, the L Gene and V H The features of the gene are input into the generation of feature vectors in the subsequent steps.

[0199] The following features can be incorporated into V L Gene and V H In the feature vector of a gene pair:

[0200] Regarding the observed V H Gene information (e.g. unique identifier)

[0201] Regarding the observed V L Gene information (e.g. unique identifier)

[0202] In the pool from the second selection cycle, the V H Absolute number of genes VH (frequency of occurrence)

[0203] V observed in the pool from the second selection cycle L Absolute number of genes VL (occurrence frequency)

[0204] In the pool from the second selection cycle, different V H Absolute number of genes numVH (occurrence frequency)

[0205] In the pool from the second selection cycle, different V L Absolute number of genes numVL (occurrence frequency)

[0206] The V that appears most frequently in the pool from the second selection cycle H Absolute number of genes maxVH (occurrence frequency)

[0207] The V that appears most frequently in the pool from the second selection cycle L Absolute number of genes maxVL (occurrence frequency)

[0208] In the pool from the second selection cycle V H The relative frequency of occurrence of the gene relVH (relative to the most frequent VH in the pool from the second selection cycle) H Number of genes): relVH = VH / maxVH

[0209] In the pool from the second selection cycle V L The relative frequency of the gene relVL (relative to the most frequent VL in the pool from the second selection cycle) L Number of genes): relVL = VL / maxVL

[0210] Observed V H The frequency of occurrence of genes in the pool from the second selection cycle is similar to the observed V L The difference diff (distance) between the frequencies of occurrence of genes in the pool from the second selection cycle (as an absolute value): diff = |VH-VL|

[0211] ·The V H The frequency of occurrence of a gene in the pool from the second selection cycle is related to the V L The relative difference relDiff (relative distance) between the frequencies of occurrence of genes in the pool from the second selection cycle (relative to the frequency of occurrence of the gene with higher frequency in the pool from the second selection cycle): relDiff = |VH-VL| / max(VH, VL), where for VH>VL, max(VH, VL) = VH and for VL≥VH, max(VH, VL) = VL

[0212] The number of selection cycles (layers) that have been completed before the second selection cycle prevnum

[0213] In the pool from the first selection cycle, the V H Absolute number of genes prevVH

[0214] In the pool from the first selection cycle, the V L Absolute number of genes prevVL

[0215] V in the pool from the first selection cycle H The frequency of occurrence of genes is related to the V in the pool from the first selection cycle. L The difference between the occurrence frequencies of genes prevDiff (distance) (as an absolute value): prevDiff = |prevVH-prevVL|

[0216] V from the first selection cycle to the second selection cycle HRelative change in gene number: prevRelDiffV H =(|VH-prevVH|) / max(VH,prevVH), where for VH>prevVH, max(VH,prevVH)=VH and for prevVH≥VH, max(VH,prevVH)=prevVH,

[0217] V from the first selection cycle to the second selection cycle L Relative change in gene number: prevRelDiffV L =(|VL-prevVL|) / max(VL,prevVL), where for VL>prevVL, max(VL,prevVL)=VL and for prevVL≥VL, max(VL,prevVL)=prevVL.

[0218] It is conceivable that in addition to the features mentioned, further information is input into the generation of the feature vector.

[0219] The feature vector is fed into the model. The model computes, for each feature vector fed into the model, the V L Gene and V H This pair of genes encodes the variable regions of the light chain and the heavy chain belonging to the same antibody and / or antibody fragment or not belonging to the same antibody and / or antibody fragment.

[0220] The model can be, for example, a classification model. Such a classification model is based on each pair of V L Gene and V H The feature vector of the gene assigns the pair to one of at least two categories. The first category includes those pairs that encode the variable regions of the light chain and the heavy chain that belong to the same antibody and / or the same antibody fragment. The second category includes those pairs that encode the variable regions of the light chain and the heavy chain that do not belong to the same antibody and / or the same antibody fragment.

[0221] In short, the classification model gives the L Gene and V H Information about whether the genes match each other. L The gene encodes the variable region of the light chain of an antibody and / or antibody fragment and V H When the genes encode the variable regions of the heavy chains of the same antibody and / or the same antibody fragment, they match each other. L Gene and V H Gene pairs are also called V L -V H Yes. When V LThe gene encodes the variable region of the light chain of an antibody and / or antibody fragment and V H When the genes encode the variable region of the heavy chain of another antibody and / or another antibody fragment, they do not match each other.

[0222] The model may be, for example, a regression model. The regression model may be based on each pair of V L Gene and V H The gene feature vector is used to calculate the V L Gene and V H The probability that a gene pair encodes the variable regions of the light chain and the heavy chain belonging to the same antibody and / or the same antibody fragment. L Gene and V H When the gene pair encodes the variable regions of the light chain and the heavy chain belonging to the same antibody and / or the same antibody fragment, the calculation result of the regression model can be, for example, 0; L Gene and V H When the gene pair encodes the variable regions of the light chain and the heavy chain belonging to the same antibody and / or the same antibody fragment, the calculated result of the regression model can be, for example, 1 or 100%. L Gene and V H For gene pairs, the calculated probability is between 0 and 1 or between 0 and 100%.

[0223] The model (eg, a classification model or a regression model) is preferably generated based on a self-learning algorithm. The model is particularly preferably generated by means of supervised learning.

[0224] Model building can be done, for example, using known antibodies and / or antibody fragments or using V sequences that are known to be compatible with each other. L Gene and V H The model can be trained using these data (training dataset).

[0225] There are many methods for building classification models, such as random forests or gradient boosting. There are also many methods for building regression models, such as logistic regression. These and other methods for classification and regression are well described in the prior art (see, for example, Norman Matloff: Statistical Regression and Classification-From Linear Models to Machine Learning, Texts in Statistical Science, CRC Press 2017, ISBN 978-1-4987-1091-6; Pratap Dangeti, Statistics for Machine Learning, Packt Publishing 2017, ISBN 978-1-78829-575-8).

[0226] The result of model building is a model (e.g., a classification model or a regression model) that can also be applied to the V of unknown antibodies and / or antibody fragments. L and V H The more similar the training dataset is to the test dataset, the higher the accuracy of the model. For example, the accuracy is high when the same matrix (antigen) is used for training and testing, and low when different matrices are used.

[0227] That is, for any V L Gene and V H Gene pairs can be predicted based on the model whether they match each other (with a limited probability). This information can be output in the next step. The output can be carried out on a display screen of a computer system, for example. This information can also be printed out via a printer or stored in a data memory.

Claims

1. A method for selecting an antibody or antibody fragment of interest from a pool derived from a selection method comprising a plurality of selection cycles, the method comprising the following steps: -Receive involves V L Gene and V H Gene features, where V L Gene and V H genes encoding the variable regions of the light and heavy chains of antibodies and / or antibody fragments from the pool resulting from said selection method, - For these pools V L Gene and V H Gene pairs are used to construct feature vectors, where each pair of V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features: •V L The frequency of occurrence of a gene in the corresponding pool, •V H The frequency of occurrence of a gene in the corresponding pool, •V L The frequency of the gene in the previous pool, •V H The frequency of the gene in the previous pool, •V H The frequency of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools, •V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool, - Identify the enrichment patterns in these feature vectors of these pools, where the enrichment pattern represents the enrichment patterns in the selection method using the corresponding V L Gene and V H The strength of gene enrichment for the antibody and / or antibody fragment, - Based on this enrichment pattern, V L Gene and V H The method further comprises the steps of: -Based on the corresponding feature vector for V in a pool L Gene and V H The gene pairs are subjected to principal component analysis and these principal components are obtained. -For each pair of V L Gene and V H Gene: A vector between a start point and an end point is obtained in the vector space constructed by principal component analysis, where the start point represents a pair of principal components for the first pool of the selection method and the end point represents the same pair of principal components for the last pool of the selection method, - calculate the length of the vector, or calculate the length of the projection of the vector on an axis of the vector space, or shift the starting point of the vector to the origin of the vector space and calculate the product of the optionally weighted principal components of the end point of the vector, and use the result of the corresponding calculation as the value for V L Gene and V H The score of the gene pair, - obtaining a ranked list of VL gene and VH gene pairs based on these score values, - Output these score values ​​and / or ranked lists to the user.

2. The method of claim 1, wherein identifying the enrichment pattern comprises the following sub-steps: - Execute V in at least three consecutive pools L Gene and V H The dimension of all feature vectors of gene pairs is reduced, and V in the corresponding pool is obtained at the same time L Gene and V H Factors of gene pairs, - By quantifying the change in the coordinate values ​​of these factors along successive cells, V L Gene and V H The score value of the gene pair.

3. The method according to any one of claims 1 or 2, comprising the following steps: -Receive involves V L Gene and V H Gene features, where V L Gene and V H The genes encode the variable regions of the light chain and the heavy chain of antibodies and / or antibody fragments, wherein the antibodies and / or antibody fragments are derived from a selection method, wherein the selection method comprises a plurality of selection cycles followed one after another, wherein the result of each selection cycle is a pool of antibodies and / or antibody fragments, -Based on these characteristics, for V in these pools L Gene and V H Gene pairs are used to construct feature vectors, where each pair of V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features: •V L The frequency of occurrence of a gene in the corresponding pool, •V H The frequency of occurrence of a gene in the corresponding pool, •V L The frequency of the gene in the previous pool, •V H The frequency of the gene in the previous pool, •V H The frequency of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools, •V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool, -Based on these feature vectors, we can get these V in a pool L Gene and V H factors of gene pairs, where these factors describe the characteristics of these pairs in a pool with fewer variables than eigenvectors, - Based on these factors, V L Gene and V H The score of the gene pair, - obtaining a ranked list of VL gene and VH gene pairs based on these score values, - Output these score values ​​and / or ranked lists to the user.

4. The method according to any one of claims 1 or 2, comprising the following steps: -Receive involves V L Gene and V H Gene features, where V L Gene and V H The genes encode the variable regions of the light chain and the heavy chain of antibodies and / or antibody fragments, wherein the antibodies and / or antibody fragments are derived from a selection method, wherein the selection method comprises a plurality of selection cycles followed one after another, wherein the result of each selection cycle is a pool of antibodies and / or antibody fragments, -Based on these characteristics, for V in these pools L Gene and V H Gene pairs are used to construct feature vectors, where each pair of V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features: •V L The frequency of occurrence of a gene in the corresponding pool, •V H The frequency of occurrence of a gene in the corresponding pool, •V L The frequency of the gene in the previous pool, •V H The frequency of the gene in the previous pool, •V H The frequency of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools, •V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool, -Based on the corresponding feature vector for V in a pool L Gene and V H The gene pairs are subjected to principal component analysis and these principal components are obtained. - Starting from the first pool of the selection method to the last pool of the selection method, for each pair of V L Gene and V H The coordinates of the principal components of the gene are added, and the length of the vector generated by adding the coordinates of these principal components is calculated, and the length of this vector is used as V L Gene and V H The score of the gene pair, -Based on these score values, you can get V L Gene and V H A ranked list of gene pairs, - Output these score values ​​and / or ranked lists to the user.

5. The method according to claim 1 , wherein one or more further features selected from the following list are input when constructing the feature vectors: • In the observed pools, different V H Absolute number of genes: numVH • In the observed pools, different V L Absolute number of genes: numVL • The V that appears most frequently in the observed pool H Absolute number of genes: maxVH • The V that appears most frequently in the observed pool L Absolute number of genes: maxVL • In the observed pool V H The relative frequency of a gene relative to the most frequent V gene in the observed pool. H Number of genes: relVH = VH / maxVH • In the observed pool V L The relative frequency of a gene relative to the most frequent V gene in the observed pool. L Number of genes: relVL = VL / maxVL • In the observed pool V H The frequency of occurrence of a gene is related to the V L The relative difference between the frequencies of genes, relative to the frequency of the gene that occurs more frequently in the observed pool: relative distance reldiff = |VH - VL| / max(VH, VL), where max(VH, VL) = VH for VH > VL and max(VH,VL) = VL for VL ≥ VH, • The number of selection cycles that have completed before the observed selection cycle: prevnum •V H Normalized frequency of occurrence of a gene in the observed pool: rpmVH = (VH / sumVH) * 1000000, where sumVH is the total number of VHs in the observed pool. H The sum of the gene occurrence frequencies VH, •V L Normalized frequency of occurrence of a gene in the observed pool: rpmVL = (VL / sumVL) * 1000000, where sumVL is the total number of VL in the observed pool. L The sum of the gene occurrence frequencies VL, •V H Normalized frequency of occurrence of genes in the previous pool: prevRpmVH = (prevVH / prevSumVH) *1000000, where prevSumVH is the normalized frequency of occurrence of all V genes in the previous pool. H The sum of the gene occurrence frequencies VH, •V L Normalized frequency of occurrence of genes in the previous pool: prevRpmVL = (prevVL / prevSumVL) *1000000, where prevSumVL is the normalized frequency of occurrence of all V genes in the previous pool. L The sum of the gene occurrence frequencies VL, • V from the first selection cycle to the second selection cycle H Relative change in gene number: prevRelDiffV H = (|VH -prevVH|) / max(VH, prevVH), where max(VH, prevVH) = VH for VH > prevVH and max(VH, prevVH) = prevVH for prevVH ≥ VH, • V from the first selection cycle to the second selection cycle L Relative change in gene number: prevRelDiffV L = (|VL -prevVL|) / max(VL, prevVL), where for VL > prevVL, max(VL, prevVL) = VL and for prevVL ≥ VL, max(VL, prevVL) = prevVL.

6. The method according to claim 1 , wherein the feature vectors include the following features: • V L Normalized frequency of occurrence of a gene in the observed pool: rpmVL = (VL / sumVL) * 1000000, where sumVL is the total number of VL in the observed pool. L The sum of the gene occurrence frequencies VL, •V H Normalized frequency of occurrence of a gene in the observed pool: rpmVH = (VH / sumVH) * 1000000, where sumVH is the total number of VHs in the observed pool. H The sum of the gene occurrence frequencies VH, •V H The frequency of genes in the corresponding pool is related to V L The absolute difference between the occurrence frequencies of genes in the corresponding pools is diff, diff = |VH - VL|, • In the observed pool V H The frequency of occurrence of a gene is related to the V L The relative difference between the frequencies of genes, relative to the frequency of the gene that occurs more frequently in the observed pool: relative distance reldiff = |VH - VL| / max(VH, VL), where max(VH, VL) = VH for VH > VL and max(VH,VL) = VL for VL ≥ VH, •V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the gene's frequency of occurrence in the previous pool (prevDiff, prevDiff = |prevVH - prevVL|), •V H Normalized frequency of occurrence of genes in the previous pool: prevRpmVH = (prevVH / prevSumVH) *1000000, where prevSumVH is the normalized frequency of occurrence of all V genes in the previous pool. H The sum of the gene occurrence frequencies VH, •V L Normalized frequency of occurrence of genes in the previous pool: prevRpmVL = (prevVL / prevSumVL) *1000000, where prevSumVL is the normalized frequency of occurrence of all V genes in the previous pool. L The sum of the gene occurrence frequencies VL, • The number of selection cycles that have completed before the observed selection cycle: prevnum • The observed pool V H The normalized frequency of genes is the same as the V in the previous pool. H The base 2 logarithm of the ratio of the normalized frequency of occurrence of genes: logRpmVH = log2(rpmVH / prevRpmsVH) • The observed pool V L The normalized frequency of genes is the same as the V in the previous pool. L The base 2 logarithm of the ratio of the normalized frequency of occurrence of genes: logRpmVL = log2(rpmVL / prevRpmsVL).

7. The method according to claim 1 or 2, wherein for each pair of V L Gene and V H At least two score values ​​are calculated for the gene, a first score value for enrichment against a murine target and a second score value for a human target, wherein for each of the at least two score values ​​an own ranking is obtained, wherein these score values ​​and / or these rankings are output to a user.

8. A system for selecting an antibody or antibody fragment of interest from a pool derived from a selection method comprising a plurality of selection cycles, the system comprising: Input unit, Feature vector generation unit, Enrichment analysis unit, Fractional value calculation unit, Sorting unit, and Output unit, - wherein the input unit is configured to obtain a voltage related to V L Gene and V H Gene characteristics, V L Gene and V H genes encoding the variable regions of the light and heavy chains of antibodies and / or antibody fragments from the pool resulting from said selection method, wherein the feature vector generating unit is configured to generate V in the pool based on the obtained features L Gene and V H The feature vector of gene pairs, where V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features: •V L The frequency of occurrence of a gene in the corresponding pool, •V H The frequency of occurrence of a gene in the corresponding pool, •V L The frequency of the gene in the previous pool, •V H The frequency of the gene in the previous pool, •V H The frequency of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools, •V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool, - wherein the enrichment analysis unit is configured to identify an enrichment pattern in the feature vector of the pool, wherein the enrichment pattern represents the enrichment pattern of the pool in the selection method using the corresponding V L Gene and V H The strength of gene enrichment for the antibody and / or antibody fragment, - wherein the score value calculation unit is configured to calculate V based on the enrichment pattern L Gene and V H The scores of the gene pairs are associated with the corresponding enrichment strengths, and the calculation also includes: -Based on the corresponding feature vector for V in a pool L Gene and V H The gene pairs are subjected to principal component analysis and these principal components are obtained. -For each pair of V L Gene and V H Gene: A vector between a start point and an end point is obtained in the vector space constructed by principal component analysis, where the start point represents a pair of principal components for the first pool of the selection method and the end point represents the same pair of principal components for the last pool of the selection method, - calculate the length of the vector, or calculate the length of the projection of the vector on an axis of the vector space, or shift the starting point of the vector to the origin of the vector space and calculate the product of the optionally weighted principal components of the end point of the vector, and use the result of the corresponding calculation as the value for V L Gene and V H Gene pair scores, - based on these scores, a ranked list of VL and VH gene pairs is obtained, - wherein the output unit is configured to output the score values ​​and / or the ranked lists to a user.

9. A computer program product for selecting an antibody or antibody fragment of interest from a pool resulting from a selection method comprising a plurality of selection cycles, comprising a data carrier and a program code which is stored on the data carrier and which, when loaded into a working memory of a computer system, causes the computer system to perform the following steps: -Receive involves V L Gene and V H Gene features, where V L Gene and V H genes encoding the variable regions of the light and heavy chains of antibodies and / or antibody fragments from the pool resulting from said selection method, - For these pools V L Gene and V H Gene pairs are used to construct feature vectors, where each pair of V L Gene and V H The genes encode variable regions belonging to the same antibody and / or antibody fragment, and each feature vector of a pool includes at least the following features: • V L The frequency of occurrence of a gene in the corresponding pool, •V H The frequency of occurrence of a gene in the corresponding pool, •V L The frequency of the gene in the previous pool, •V H The frequency of the gene in the previous pool, •V H The frequency of genes in the corresponding pool is related to V L The absolute difference between the frequencies of genes in corresponding pools, •V H The frequency of gene occurrence in the previous pool is related to V L The absolute difference between the frequency of occurrence of a gene in the previous pool, - Identify the enrichment patterns in these feature vectors of these pools, where the enrichment pattern represents the enrichment patterns in the selection method using the corresponding V L Gene and V H The strength of gene enrichment for the antibody and / or antibody fragment, - Based on this enrichment pattern, V L Gene and V H The method further comprises the steps of: -Based on the corresponding feature vector for V in a pool L Gene and V H The gene pairs are subjected to principal component analysis and these principal components are obtained. -For each pair of V L Gene and V H Gene: A vector between a start point and an end point is obtained in the vector space constructed by principal component analysis, where the start point represents a pair of principal components for the first pool of the selection method and the end point represents the same pair of principal components for the last pool of the selection method, - calculate the length of the vector, or calculate the length of the projection of the vector on an axis of the vector space, or shift the starting point of the vector to the origin of the vector space and calculate the product of the optionally weighted principal components of the end point of the vector, and use the result of the corresponding calculation as the value for V L Gene and V H The score of the gene pair, - obtaining a ranked list of VL gene and VH gene pairs based on these score values, - outputting these score values ​​and / or the ranked list to a user.

10. The computer program product according to claim 9, wherein the program code causes the computer system to perform one or more steps of the method recited in claims 1 to 7.

Images