Next Article in Journal
Diversity of Neurotransmitter-Producing Human Skin Commensals
Previous Article in Journal
Molecular Mechanisms of Heterosis and Its Applications in Tree Breeding: Progress and Perspectives
Previous Article in Special Issue
The Trajectory of Damaged-Base Eversion into the Active Site of Apurinic/Apyrimidinic Endonuclease APE1 Regulates This Enzyme’s Substrate Specificity
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
IJMS
Volume 25
Issue 22
10.3390/ijms252212343
ijms-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Similarity Analysis of Computer-Generated and Commercial Libraries for Targeted Biocompatible Coded Amino Acid Replacement

by
Markus Meringer
1,
Gerardo M. Casanola-Martin
2,
Bakhtiyor Rasulev
2,3 and
H. James Cleaves II
4,5,6,*
1
German Aerospace Center (DLR), Department of Atmospheric Processors, Oberpfaffenhofen, 82234 Wessling, Germany
2
Department of Coatings and Polymeric Materials, North Dakota State University, Fargo, ND 58108, USA
3
Department of Chemistry, National University of Uzbekistan, Tashkent 100174, Uzbekistan
4
Department of Chemistry, Howard University, Washington, DC 20059, USA
5
Earth-Life Science Institute, Tokyo Institute of Technology, 2-12-1-IE-1 Ookayama, Meguro-ku, Tokyo 152-8550, Japan
6
Blue Marble Space Institute for Science, 1001 4th Ave, Suite 3201, Seattle, WA 98154, USA
*
Author to whom correspondence should be addressed.
Int. J. Mol. Sci. 2024, 25(22), 12343; https://doi.org/10.3390/ijms252212343
Submission received: 25 October 2024 / Revised: 10 November 2024 / Accepted: 12 November 2024 / Published: 17 November 2024
(This article belongs to the Special Issue Computational Approaches for the Investigation of Complex Molecular Mechanisms, 3rd Edition)
Browse Figures
Versions Notes

Abstract

:
Many non-natural amino acids can be incorporated by biological systems into coded functional peptides and proteins. For such incorporations to be effective, they must not only be compatible with the desired function but also evade various biochemical error-checking mechanisms. The underlying molecular mechanisms are complex, and this problem has been approached previously largely by expert perception of isomer compatibility, followed by empirical study. However, the number of amino acids that might be incorporable by the biological coding machinery may be too large to survey efficiently using such an intuitive approach. We introduce here a workflow for searching real and computed non-natural amino acid libraries for biosimilar amino acids which may be incorporable into coded proteins with minimal unintended disturbance of function. This workflow was also applied to molecules which have been previously benchmarked for their compatibility with the biological translation apparatus, as well as commercial catalogs. We report the results of scoring their contents based on fingerprint similarity via Tanimoto coefficients. These similarity scoring methods reveal candidate amino acids which could be substitutable into modern proteins. Our analysis discovers some already-implemented substitutions, but also suggests many novel ones.

1. Introduction

Modern biology uses 20 canonical coded amino acids (CAAs); however, there are occasional coded substitutions (e.g., selenocysteine and pyrrololysine [1], and many other non-canonical amino acids have been engineered or otherwise substituted into proteins [2]. It is believed that early organisms used simpler and more promiscuously coded amino acid sets, and that the tolerance of biological proteins and the processes that makes them into dissimilar coded amino acids has changed over time [3,4,5]. Since tRNA amino acid charging is accomplished and error-checked by proteins, this reduction in ambiguity occurred in tandem with the development of the biological machinery which accomplishes it. This machinery developed in the context of the compounds present in evolving organisms, and there are a great number of possible isomers of any given amino acid, many of which are real metabolites. Indeed, organisms have evolved various secondary metabolite amino acid analogues which can interfere with protein biosynthesis in predators or competitors and are not recognized by their own translation machinery [6].
Proteins, and other biological macromolecules, can tolerate multiple xenobiological monomer substitutions (see, for example, Marlière et al. [7], Malyshev et al. [8]). This is mainly due to the way in which biological information is communicated in molecules, e.g., a combination of shape and surface electron density interactions between macromolecules and small molecules governs the strength of molecular interaction [9]. For this reason, drugs are effective modulators of enzymatic activity, and molecular similarity is a powerful tool for finding them [10].
Biological information encoding, which entails the reliable connection of structural interaction and kinetic information with downstream biological “meaning,” is mediated by several processes involved in translation, including the interactions of amino acids and aminoacyl-tRNA synthetases during tRNA charging and the interactions of aminoacyl-tRNAs with the ribosome during translation. Furthermore, as it is likely that the genetic code has evolved over time [3,5], this interfacial “language” has had to adapt to the ways that the incorporation of new structural elements enabled new coding dynamics over time.
Organic chemical structure space is vast [11]. This has implications for why biology uses the compounds it does [12]. Biology’s choice of components is directed and constrained by its composition at any given point in time [13,14]. This is analogous to the development of languages and syntaxes: initially poorly constrained relational systems which interact by feedback with mechanisms which enforce their existence can rapidly reinforce canonical encodings [15]. However, such analyses are likely blind to interfering noise which does not exist in the learning system during its development. Such noise can be introduced later into such systems and potentially tolerated well. This dynamic likely lies at the heart of biological evolution.
As an example of a non-engineered incorporation of a non-canonical amino acid into biological proteins, it has been shown that norleucine (norleu) can substitute for methionine (M) in proteins due to their molecular similarity [2]. The CH2CH2CH3 structural motif in norleu is apparently similar enough to M’s CH2SCH3 motif that this compound evades protein synthesis error-correction mechanisms and becomes incorporated in biological proteins using the same coding mechanisms. However, norleu is not a major intermediate in any known biosynthetic pathway, and thus, most cells never experience situations in which they must carefully distinguish between M and norleu. In this case, norleucine may be functionally substitutable, tolerably substitutable with reduced function, or a complete poison for any given protein, but this situation rarely arises.
Compounds which are structurally compatible because their functional groups do not interfere with normal biological interactions are known as bioisosteres [16]. Bioisosteres may be rather common in chemical space; this study aims to provide a novel method to find them.
Amino acids are especially attractive targets for bioisostere exploration because of their fundamental roles in biochemistry. In addition to amino acid analogues which may serve as enzyme agonists or antagonists, coded amino acid analogues can modulate coded protein function [17], and amino acids are important intermediates in many biosynthetic pathways as well as important neurotransmitters (e.g., GABA, glutamic acid (E), aspartic acid (D), serine (S), glycine (G) [18]). There have been a few reported screens examining the promiscuity of the translation apparatus, which examined scores of purchasable amino acids, including ɑ-, β-, and ɑ,ɑ-substituted amino acids [19,20].
Introducing new monomers in biopolymers often depends on finding compounds with similar shape and electrotopological properties. This could be especially informative in the case of coded amino acids used in biopolymers which may have multiple levels at which their compatibility is proofread by biological processes. For amino acid incorporation in proteins, amino acids must be recognized by tRNA synthetases, then be tolerated by the ribosomal machinery, and finally not interfere too strongly with the resulting protein’s folding and function. While incorporation of a single residue stochastically at any given site in a protein sequence may be tolerable, multiple or complete substitution may not be. Some sites may also be more sensitive to substitutions than others, or particular types of substitutions may be problematic for many reasons. In the case of amino acids present in the growth medium, there must also be mechanisms for importation of the molecules inside the cell. Thus, substitution can fail at many levels.
There are multiple computational methods already developed for performing such searche [21,22,23]. Our approach is novel in its use of the combination of structure enumeration and similarity metrics, and its application to amino acid structures. The similarity and biocompatibility of amino acids are important to analyze to understand possible prebiotic interactions between peptides and other molecules. In a previous study [24], we generated virtual libraries of amino acids using MOLGEN 5.0 structure generation software [25,26]. These sets were designed starting from molecular formulas and sub-formulas of the 20 CAAs using lists of implausible substructures to exclude sterically and energetically unstable molecules. This resulted in a virtual library surrounding the chemical space defined by the 20 CAAs. This dataset was primarily developed for exploration of the evolution of the genetically encoded amino acids [13,14,27,28]) and we use it here to search for potentially biocompatible amino acids.

2. Results and Discussion

2.1. Composition of the Library

After removal of duplicate structures, applying the procedure described in the Section 3, we arrived at a total of 11,302 structures. Table 1 contains the sizes of each set, and Figure 1 shows a Venn diagram of the overlap of EAAs, GAAs, and PAAs, which underscores the notion that this space has not been extensively explored to date. The SI contains a comma-separated value (.csv) file with SMILES representations of all 11,302 AAs in the rows and columns indicating the set memberships of each structure.

2.2. Tanimoto Coefficients

TCs are stored in a square matrix of 11,032 rows and columns, which we call the similarity matrix. This matrix is symmetric because, for two AAs X and Y, we have TC(X, Y) = TC(Y, X), and there are entries of 1 on the main diagonal, because TC(X, X) = 1. In Figure 2, which shows the block structure of the similarity matrix, this is depicted by entries of 1 on the main diagonal, and arbitrary entries in the upper triangle matrix represented by asterisks. We will focus on this triangle matrix to avoid counting TCs of the same pair of AAs twice. This triangle matrix has 11,302 ∙ 11,301/2 = 63,861,951 entries. The black stroke curve of Figure 3 shows cumulative relative frequencies of these entries; Figure S1 (bottom left) of the SI shows the distribution as a histogram, and the last column of Table 2 reports key statistical parameters for these TCs.
Focusing on the CAA block of the similarity matrix, the upper triangle matrix has 20 ∙ 19/2 = 190 entries. These entries are highlighted in red in Figure 2; cumulative relative frequencies are depicted in the red curve in Figure 3. A histogram of these TCs among pairs of CAAs is shown in the top left panel of Figure S1, and key statistical parameters are reported in column CAA of Table 2.
Of particular interest for this study is the top right block of the similarity matrix, with entries colored in turquoise in Figure 2. This block contains all TCs of pairs composed of a CAA and an XAA. It has 20 ∙ 11,282 = 225,640 entries; cumulative relative frequencies are depicted by the turquoise curve in Figure 3, and histograms are shown in the right panels of Figure S1. We refer to these TCs as CAA × XAA, denoting that here, the similarity for all pairs of the Cartesian product of CAA and XAA are covered.
Comparing the cumulative relative frequencies of Figure 3, we notice that TCs of AAAs are tendentially slightly lower than those of CAA × XAA, while those of CAAs are in turn somewhat higher than those of CAA × XAA. This can, as well, be seen by comparing the histograms of Figure S1, and is also numerically confirmed in Table 2. The first and the third quartiles, as well as the reported quantiles, median, and mean are lowest for AAA and highest for CAA TCs. It is also interesting to note that the AAA TCs have a maximum value of 1, meaning there must be pairs of structures with identical fingerprints among our AAs (cf. SI section Choice of Fingerprints and Figure S2).
Table 3 shows a detailed view of all TCs for CAAs. Columns and rows are sorted with respect to increasing molecular weight and labeled with one-letter CAA abbreviations. Table cells are color-coded by a red–yellow–green scale, with red corresponding to low, and green to high, TC. The red colors in the row and column P indicate the low similarity of proline with all other CAAs, which can be explained by P’s unique cyclic amino acid backbone. Looking for further regions of low similarity, we see that the low-weight AAs G, alanine (A), valine (V), and threonine (T) have low TC, with the high-weight AAs histidine (H), phenylalanine (F), arginine (R), tyrosine (Y), and tryptophan (W), resulting in orange-colored, rectangular-shaped areas in the upper right and lower left corners of the CAA similarity matrix. This pattern is only interrupted by some yellow-colored cells in the row and column of serine (S). The maximum TC of 0.65 among CAAs is met by V and Y. We will use this value of 0.65 later as a threshold for presenting the most similar XAAs for each CAA. Further highly similar pairs of CAAs are A and V, with TC(A, V) = 0.611; asparagine (N) and aspartic acid (D), with TC(N, D) = 0.636; and glutamine (Q) and glutamic acid (E), with TC(Q, E) = 0.607. Figure S3 shows cumulative relative frequencies of TCs for each of the 20 CAAs as color-coded curves, and in SI section TC Outliers and Choice of TC Threshold, we show an outlier analysis of TCs of CAA × XAA with respect to the 20 CAAs using box plots (Figure S4). We find that there are a total of 566 extreme outliers at the high TC end, which would be promising candidates for experimental evaluation. Below, we identify structures of the highest similarity with each CAA.

2.3. Ranking

Using the Tanimoto coefficients, we can sort the structures corresponding to decreasing similarity for each CAA. Figure 4 shows a selection of highly ranked, i.e., very similar, XAAs for each CAA. The following selection rules were applied:
  • Report for each CAA the most similar XAA.
  • Show further highly ranked XAAs for each CAA with TC greater than 0.65.
For more reasoning behind these rules, the reader is referred to the SI section TC Outliers and Choice of TC Threshold.
The structures of Figure 4 have labels composed of a one-letter CAA abbreviation and one or two numbers, e.g., G1 is the highest ranked XAA for glycine and correspondingly A1 for alanine. Table 4 lists, for each CAA, the highest TCs, the corresponding XAA (or CAA) with its abbreviation (ID) from Figure 4, and its set memberships, e.g., G1 has a rather low TC of 0.45 with G. It is included in GAAs, but not in EAAs nor PAAs. For alanine, the highest TC of 0.611 is achieved twice, by A1 and V (omitted in Figure 3, as the CAA structures are generally known). For serine, we again have just a unique highest ranked XAA, S1, with a TC of 0.6. Proline has several analogs with rather high TCs (see Figure S4). The three highest ranked XAAs are labeled P1, P2, and P3 in Figure 4. For V, we have four AAs with TC values above the threshold of 0.65. The highest-ranked structure is labeled, as usual, with V1. The second highest-ranked is identical with A1, i.e., label V2 does not appear in Figure 4 or Table 4. The next highest-ranked structures both have the same TC of 0.667 and thus have labels V3.1 and V3.2 to indicate that both achieve rank three. For T, we have, three times, TC above 0.65, achieved by T1 and the equally scoring T2.1 and T2.2. For cysteine (C), the highest TC is 0.619, i.e., below the threshold of 0.65. Thus, only the highest ranked XAA for C, C1, is depicted in Figure 4. The highest-ranked structure for isoleucine (I) is I1, with a TC of 0.667, and there are no further AAs with TC above 0.65. The most similar AA for leucine (L) is L1, with a TC of 0.643. The highest-ranked structure for N is D and vice versa. We therefore depict the second highest-ranked D2 and N2 in Figure 4. The situation is similar with Q and E, leading to Q2 and E2 shown in Figure 4. The highest-ranked structure for lysine (K) is K1, with a rather high TC of 0.714, and there are no further similar AAs for K with TC above 0.65. The highest-ranked structures for M, H, and F are all below the threshold, and are labeled M1, H1, and F1. For R, the four highest TCs are above 0.65, and the corresponding structures are shown as R1, R2, R3, and R4 in Figure 4. Finally, for Y and W, the highest-ranked AAs again have TCs below 0.65, and the corresponding structures have labels Y1 and W1 in Figure 4.
We note that ornithine (K1) and 1, 3-diaminopropionic acid (S1) are, meanwhile, known to be poor bioisosteres (see e.g., Frenkel-Pinter et al. [29], and Makarov et al. [30]). However, it is not the subject of the present study to review negative experimental results on potential bioisosteres, and we refer to [31] for a comprehensive review of non-canonical tRNA synthetase substrates.
We see in Table 4 that the vast majority of this small selection is commercially available and has not yet been studied experimentally. The SI contains TCs for each CAA, with the entire library as a .CSV file. Further suggestions on how to evaluate AAs with respect to decreasing similarity to a given CAA, and to visualize the most similar analogs, are also offered in the SI. This gives access to a potentially rich set of easily accessible AAs, which may be promising candidates for experimental evaluation.

2.4. Projection

Figure 5 shows a 2D projection of our AA chemical space using classical MDS with Tanimoto distances as input. We see the smallest CAA, G, taking a central position in this mapping. P is located in the top right region, quite distant from all other CAAs. This again reflects the special structure of P, which we have already noted before when inspecting the matrix of TCs for the CAAs in Table 3. On the left, we see the group of the four aromatic AAs marked by an ellipse. F and Y are located very close to each other, with H also nearby, and W, the largest CAA, a bit further away, at the top left corner of this group. The remaining 14 CAAs are located close to each other in an area marked by a rectangle in Figure 5. The inset at the bottom left corner zooms into this densely populated area. Here, we see closely located pairs of CAAs, such as V and Y (their high TC was mentioned above), C and S (which differ only in their OH/SH group), or the pair of isomers L and I. However, we note that such a projection cannot reflect all similarity relationships perfectly. For instance, N is the most similar CAA for D, but appears at a relatively large Euclidean distance to D in this figure, whereas R, with its rather low TC(R, D) = 0.316, has the shortest Euclidean distance to D among all CAAs in this MDS. There are multiple reasons for these imperfections, particularly the fact that this MDS is calculated to optimize Euclidean distances for all pairs of AAAs, and CAAs are only a tiny subset of AAAs. We present a deeper analysis of the MDS quality in the corresponding SI section. Despite these imperfections, this 2D projection offers a useful alternative to the 1D approach realized by ranking with respect to single CAAs. For instance, using this method, XAAs can be discovered that are not among the top-ranked ones by using the 1D approach, but which are nevertheless interesting due to relatively high simultaneous similarities with multiple CAAs. Another advantage of the 2D projection could be the discovery based on diversity, and to inspect regions not yet covered by EAAs. To simplify access to the structures of XAAs in this projection, we included their 2D coordinates in the .csv file (see SI file AASS.csv). Further recommendations on software to open this .csv file, reproduce the 2D layout, zoom into interesting regions, and display selected structures of interest are also given at the end of the SI section on Detailed Results. The quality of our 2D layout is further discussed in the SI section MDS Quality and Tables S1–S3 included there.

3. Materials and Methods

The set of computationally Generated Amino Acid (GAA) structures used in the present study contains 1913 real and hypothetical compounds, including the 20 CAAs of the standard genetic code, as described in [13]. Additionally, we imported the amino acids tested by Hartman et al. [20] and Josephson et al. [19] via their annotated Chemical Abstracts Service (CAS) number entries as listed on SciFinder Scholar. We refer to this set as Experimentally evaluated Amino Acids (EAAs). We also downloaded amino analogue structures with the HNCCOOH substructure motif under 200 amu from Emolecules (https://www.emolecules.com/ accessed on 2 April 2024) as an .sdf file, and we call this set Purchasable Amino Acids (PAAs).
The further processing is sketched in Figure 6. In a standardization step, the PAA set was cleaned of salts, and stereochemical information was removed using the KNIME [32] Chemical Identifier Resolver (CIR) node (AlvaScience, Lecco, Italy). Then, all sets were converted to canonical SMILES using OpenBabel [33]. These were copied into one file and sorted lexicographically, and duplicates were removed. Set memberships of each compound were stored in a table with five columns—SMILES—and columns representing the four sets CAA, EAA, GAA, and PAA. Boolean entries true or false in these columns indicate whether a compound belongs to the set. For this step and all forthcoming computations, we used R version 4.3.1 [34].
Compounds were scored for structural similarity using 2D molecular fingerprints and Tanimoto Coefficients (TCs). For compound representation, we computed Extended Connectivity FingerPrints of diameter 6 (ECFP6s), [35]. ECFP6s are circular fingerprints that encode atom-type information and bond connectivity up to a depth of six non-hydrogen atoms, intending to capture precise atom environment substructural features. The reason for selecting diameter 6 is exposited in the Supplementary Information (SI) section Choice of Fingerprints. Fingerprints were computed using the rcdk package (version 3.8.1, [36]) starting from SMILES-level representations and resulting in 1024-bit bit-vectors.
TCs [37] were computed among the fingerprint representations with the same R package. They were used as similarity scores and stored in a symmetrical matrix called the similarity matrix. For more mathematical background on TCs, the reader is referred to the SI section Tanimoto Coefficients. We sorted the data such that the first rows and columns correspond to CAAs and the latter to non-canonical AAs, which we call Xeno Amino Acids (XAAs). This imposes a block structure to the similarity matrix, which is discussed in detail in Section 2.2. Different blocks of the similarity matrix representing TCs among CAAs, TCs of CAAs with XAAs, and TCs among all amino acids (AAAs) were analyzed using standard statistical parameters (minima, maxima, means, quantiles, etc.). Structures were sorted by decreasing similarity with each of the 20 CAAs. Distributions of TCs were determined using the R function ecdf (see Table S2 for the most important R functions used in this study, together with their R package names and version numbers). We conducted an outlier analysis of XAAs based on TCs with CAAs, and the highest-ranked XAAs were determined.
To visualize similarities among AAs, we used classical Multi-Dimensional Scaling (MDS, [38,39,40,41]). MDS algorithms take a distance matrix as input and calculate n-dimensional coordinates for each object, i.e., AAs, such that Euclidean distances in n-dimensional space optimally represent the input distances. We chose n = 2 to represent AAs as a 2-dimensional scatterplot. We used the R function cmdscale, which provides the method of Gower [40]. This method is also known as principal coordinate analysis [42]. To convert our similarity matrix given by TCs into a distance matrix, we used the general relationship between similarity and distance [43], i.e., similarity = 1/(1 + distance), and calculated the Tanimoto Distances (TD) as TD = (1—TC)/TC.
We note that TC is always greater than zero in our similarity matrix, so we do not have to face the problem of denominators being zero in this equation.

4. Conclusions

These results suggest that there remain numerous untested XAAs that could be substituted into contemporary proteins using the modern translation apparatus. Our original study [24] was limited to amino acids that were most similar to modern biological amino acids, for example, those containing only the elements CHNOS. Many alterations of structure not originally considered in the constructed set are possible, and more nuanced sets could be produced for this sort of comparison [44]. This can be done simply by adapting our previously developed workflow [24]); thus, the current study should be considered a preliminary proof of concept. The inclusion of purchasable AAs in this study sheds some light as to the compositional search space that could be explored (for example, by inclusion of other chemical elements or other formula ranges).
Biological component substitutability likely correlates with both depth of evolutionary biological embeddedness and the permittivity of structure space. For small amino acids, fewer substitutions are possible, simply because the search space is small.
This methodology can likely usefully screen bioactive compounds. Typically, in drug screening, many millions of compounds must be searched to find an active molecule [45]. The CAAs are unlike drugs “in general” in significant ways, and we have screened a much smaller set of compounds than is usually explored in modern drug screening (see for example, The Atomwise AIMS Program [46]).
Amino acids are largely synthesized by organisms themselves, though many higher-trophic-level organisms, including humans, have incomplete amino acid biosynthesis capabilities. This kind of nested interdependency is a common phenomenon among cofactors and primary metabolites [47]. The use of the CAAs likely reflects this dependency. The CAAs also share a common co-evolutionary history (possibly extending back billions of years [3,5]) that has depended on and constrained the types of molecular interactions that were allowed among them as proteins evolved. The CAAs appear to have been evolutionarily selected to be both biosynthetically and functionally “minimal” [5], meaning that they are not only among the lowest-molecular-weight compounds (and therefore have relatively few isomers, as the number of possible isomers grows exponentially with the number of atoms in a molecule [24]) that contain any given functionality, but they are also among the biosynthetically simplest and energetically least costly. Given these deep constraints, it might be surprising if there are many CAA analogues that can be found that can be easily substituted into modern proteins without violating these concepts.
The methodology presented here may also offer a way to screen for mechanisms by which biology arrived at its monomer usage, as compatible bioisosteres or otherwise electrotopologically similar compounds may not be synthetically easily reachable via feasible metabolisms. Thus, there may be an underlying structure to chemical space which guides biology to select certain molecules, e.g., once biology uses a set of compounds, it is predisposed to use similar ones and avoid dissimilar components. If this is true, we might expect biology elsewhere in the universe to be very similar to terrestrial biology, especially as there may be certain general entry points for environmentally common C, N or S-containing raw materials (e.g., CO2, CH4, N2, NH3, H2S, SO2, etc.) into the network of chemically allowed biotransformations.
The canonical AAs, if they cluster according to definable properties such as electrotopological metrics, may then represent a built-in teleological endpoint for evolution, in that the evolution of pathways for their formation also shapes their synthesis during biology’s exploration of fitness landscapes.
We note that there are possible relationships of the present study with the concept of substitution matrices [48,49], and a systematic exploration of these relationships may be a promising starting point for future work.
We presented here, for the first time, and for a very large dataset of XAAs, methodologies and results to select candidates for CAA substitutions in proteins based on structural similarity. The provided data can significantly refine the selection process of non-canonical amino acids for various applications in protein engineering (see SI section Applications of Xeno Amino Acids).

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/ijms252212343/s1. All reference [50,51,52,53,54,55,56,57,58,59] in supplementary are cited in text.

Author Contributions

Conceptualization, H.J.C.II; Methodology, M.M. and B.R.; Project administration, H.J.C.II; Software, M.M., G.M.C.-M. and B.R.; Visualization, M.M. and G.M.C.-M.; Writing—original draft, M.M. and H.J.C.II; Writing—review & editing, M.M., G.M.C.-M. and B.R. All authors have read and agreed to the published version of the manuscript.

Funding

The authors would like to thank the Earth-Life Science Institute (ELSI) for support during the preparation of this work, and ELSI Origins Network (EON) for generous travel support for MM and BR. This work was also supported by a JSPS KAKENHI Grant-in-Aid for Scientific Research on Innovative Areas “Hadean Bioscience”, Grant Number JP26106003, and by the EON, which is supported by a grant from the John Templeton Foundation. The opinions expressed in this publication are those of the author(s) and do not necessarily reflect the views of the John Templeton Foundation. This work also used resources of the Center for Computationally Assisted Science and Technology (CCAST) at North Dakota State University, which were made possible, in part, by National Science Foundation (NSF) MRI award No. 2019077.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

All data for reproducing results of our amino acid similarity study are available in the SI file AASS.csv and described in the SI section Detailed Results. All software used for this study is publicly available for free, and is cited in Section 3 of the main text (KNIME, OpenBabel, R) and in the SI (DataWarrior).

Acknowledgments

The authors would like to thank Matteo Bartolo for help with the use of the AlvaScience CIR KNIME node.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Yuan, J.; O’Donoghue, P.; Ambrogelly, A.; Gundllapalli, S.; Sherrer, R.L.; Palioura, S.; Simonović, M.; Söll, D. Distinct genetic code expansion strategies for selenocysteine and pyrrolysine are reflected in different aminoacyl-tRNA formation systems. FEBS Lett. 2010, 584, 342–349. [Google Scholar] [CrossRef]
  2. Gilles, A.M.; Marlière, P.; Rose, T.; Sarfati, R.; Longin, R.; Meier, A.; Fermandjian, S.; Monnot, M.; Cohen, G.N.; Bârzu, O. Conservative replacement of methionine by norleucine in Escherichia coli adenylate kinase. J. Biol. Chem. 1988, 263, 8204–8209. [Google Scholar] [CrossRef]
  3. Trifonov, E.N. Consensus temporal order of amino acids and evolution of the triplet code. Gene 2000, 261, 139–151. [Google Scholar] [CrossRef]
  4. Woese, C.R.; Olsen, G.J.; Ibba, M.; Söll, D. Aminoacyl-tRNA synthetases, the genetic code, and the evolutionary process. Microbiol. Mol. Biol. Rev. MMBR 2000, 64, 202–236. [Google Scholar] [CrossRef]
  5. Cleaves, H.J. The origin of the biologically coded amino acids. J. Theor. Biol. 2010, 263, 490–498. [Google Scholar] [CrossRef]
  6. Song, Y.; Zhou, H.; Vo, M.-N.; Shi, Y.; Nawaz, M.H.; Vargas-Rodriguez, O.; Diedrich, J.K.; Yates, J.R.; Kishi, S.; Musier-Forsyth, K.; et al. Double mimicry evades tRNA synthetase editing by toxic vegetable-sourced non-proteinogenic amino acid. Nat. Commun. 2017, 8, 2281. [Google Scholar] [CrossRef]
  7. Marlière, P.; Patrouix, J.; Döring, V.; Herdewijn, P.; Tricot, S.; Cruveiller, S.; Bouzon, M.; Mutzel, R. Chemical Evolution of a Bacterium’s Genome. Angew. Chem. Int. Ed. 2011, 50, 7109–7114. [Google Scholar] [CrossRef]
  8. Malyshev, D.A.; Dhami, K.; Lavergne, T.; Chen, T.; Dai, N.; Foster, J.M.; Corrêa, I.R., Jr.; Romesberg, F.E. A semi-synthetic organism with an expanded genetic alphabet. Nature 2014, 509, 385–388. [Google Scholar] [CrossRef]
  9. Kuriyan, J.; Konforti, B.; Wemmer, D. The Molecules of Life: Physical and Chemical Principles; W.W. Norton & Company: New York, NY, USA, 2012. [Google Scholar] [CrossRef]
  10. Dunn, I.S. Searching for Molecular Solutions; Wiley: Hoboken, NJ, USA, 2009. [Google Scholar]
  11. Dobson, C.M. Chemical space and biology. Nature 2004, 432, 824–828. [Google Scholar] [CrossRef]
  12. Meringer, M.; Cleaves, H.J. Exploring astrobiology using in silico molecular structure generation. Philos. Transactions. Ser. A Math. Phys. Eng. Sci. 2017, 375, 20160344. [Google Scholar] [CrossRef]
  13. Ilardo, M.; Meringer, M.; Freeland, S.; Rasulev, B.; Cleaves, H.J., 2nd. Extraordinarily adaptive properties of the genetically encoded amino acids. Sci. Rep. 2015, 5, 9414. [Google Scholar] [CrossRef] [PubMed]
  14. Ilardo, M.; Bose, R.; Meringer, M.; Rasulev, B.; Grefenstette, N.; Stephenson, J.; Freeland, S.; Gillams, R.J.; Butch, C.J.; Cleaves, H.J., 2nd. Adaptive Properties of the Genetically Encoded Amino Acid Alphabet Are Inherited from Its Subsets. Sci. Rep. 2019, 9, 12468. [Google Scholar] [CrossRef] [PubMed]
  15. Casti, J.L.; Karlqvist, A. (Eds.) Complexity, Language, and Life: Mathematical Approaches. In Biomathematics; Springer-Verlag: Berlin/Heidelberg, Germany, 1980; Volume 16. [Google Scholar]
  16. Brown, N. Bioisosteres in Medicinal Chemistry; Wiley-VCH: Weinheim, Germany, 2012; p. 238. [Google Scholar]
  17. Drienovská, I.; Roelfes, G. Expanding the enzyme universe with genetically encoded unnatural amino acids. Nat. Catal. 2020, 3, 193–202. [Google Scholar] [CrossRef]
  18. Miyajima, M. Amino acids: Key sources for immunometabolites and immunotransmitters. Int. Immunol. 2020, 32, 435–446. [Google Scholar] [CrossRef]
  19. Josephson, K.; Hartman, M.C.T.; Szostak, J.W. Ribosomal Synthesis of Unnatural Peptides. J. Am. Chem. Soc. 2005, 127, 11727–11735. [Google Scholar] [CrossRef]
  20. Hartman, M.C.T.; Josephson, K.; Lin, C.-W.; Szostak, J.W. An expanded set of amino acid analogs for the ribosomal translation of unnatural peptides. PLoS ONE 2007, 2, e972. [Google Scholar] [CrossRef]
  21. Willett, P.; Barnard, J.M.; Downs, G.M. Chemical Similarity Searching. J. Chem. Inf. Comput. Sci. 1998, 38, 983–996. [Google Scholar] [CrossRef]
  22. Bender, A.; Glen, R.C. Molecular similarity: A key technique in molecular informatics. Org. Biomol. Chem. 2004, 2, 3204. [Google Scholar] [CrossRef]
  23. Maldonado, A.G.; Doucet, J.P.; Petitjean, M.; Fan, B.-T. Molecular similarity and diversity in chemoinformatics: From theory to applications. Mol. Divers. 2006, 10, 39–79. [Google Scholar] [CrossRef]
  24. Meringer, M.; Cleaves, H.J.; Freeland, S.J. Beyond Terrestrial Biology: Charting the Chemical Universe of α-Amino Acid Structures. J. Chem. Inf. Model. 2013, 53, 2851–2862. [Google Scholar] [CrossRef]
  25. Meringer, M. Structure Enumeration and Sampling. Chapman Hall/CRC Math. Comput. Biol. 2010, 233–267. [Google Scholar] [CrossRef]
  26. Gugisch, R.; Kerber, A.; Kohnert, A.; Laue, R.; Meringer, M.; Rücker, C.; Wassermann, A. MOLGEN 5.0, A Molecular Structure Generator. In Advances in Mathematical Chemistry and Applications; Basak, S.C., Restrepo, G., Villaveces, J.L., Eds.; Bentham Science Publisher: Amsterdam, The Netherlands, 2014; Volume 1, p. 1:113. [Google Scholar]
  27. Bose, R.; Meringer, M.; Ilardo, M.; Cleaves, H.J. Adaptive properties of the amino acid alphabet and its subsets. In Artificial Life Conference Proceedings; MIT Press: Cambridge, MA, USA, 2018; pp. 459–460. [Google Scholar]
  28. Brown, S.M.; Voráček, V.; Freeland, S. What Would an Alien Amino Acid Alphabet Look Like and Why? Astrobiology 2023, 23, 536–549. [Google Scholar] [CrossRef]
  29. Frenkel-Pinter, M.; Haynes, J.W.; C, M.; Petrov, A.S.; Burcar, B.T.; Krishnamurthy, R.; Hud, N.V.; Leman, L.J.; Williams, L.D. Selective incorporation of proteinaceous over nonproteinaceous cationic amino acids in model prebiotic oligomerization reactions. Proc. Natl. Acad. Sci. USA 2019, 116, 16338–16346. [Google Scholar] [CrossRef]
  30. Makarov, M.; Sanchez Rocha, A.C.; Krystufek, R.; Cherepashuk, I.; Dzmitruk, V.; Charnavets, T.; Faustino, A.M.; Lebl, M.; Fujishima, K.; Fried, S.D.; et al. Early Selection of the Amino Acid Alphabet Was Adaptively Shaped by Biophysical Constraints of Foldability. J. Am. Chem. Soc. 2023, 145, 5320–5329. [Google Scholar] [CrossRef]
  31. Hartman, M.C.T. Non-canonical Amino Acid Substrates of E. coli Aminoacyl-tRNA Synthetases. Chembiochem A Eur. J. Chem. Biol. 2022, 23, e202100299. [Google Scholar] [CrossRef]
  32. Berthold, M.R.; Cebron, N.; Dill, F.; Gabriel, T.R.; Kötter, T.; Meinl, T.; Ohl, P.; Thiel, K.; Wiswedel, B. KNIME-the Konstanz information miner. ACM SIGKDD Explor. Newsl. 2009, 11, 26–31. [Google Scholar] [CrossRef]
  33. O’Boyle, N.M.; Banck, M.; James, C.A.; Morley, C.; Vandermeersch, T.; Hutchison, G.R. Open Babel: An open chemical toolbox. J. Cheminformatics 2011, 3, 33. [Google Scholar] [CrossRef]
  34. R CoreTeam. R: A Language and Environment for Statistical Computing. Available online: http://www.R-project.org/ (accessed on 2 April 2024).
  35. Rogers, D.; Hahn, M. Extended-Connectivity Fingerprints. J. Chem. Inf. Model. 2010, 50, 742–754. [Google Scholar] [CrossRef]
  36. Guha, R. Chemical Informatics Functionality in R. J. Stat. Softw. 2007, 18, 1–16. [Google Scholar] [CrossRef]
  37. Tanimoto, T.T. An Elementary Mathematical Theory of Classification and Prediction; International Business Machines Corporation: New York, NY, USA, 1958. [Google Scholar]
  38. Torgerson, W.S. Multidimensional scaling: I. Theory and method. Psychometrika 1952, 17, 401–419. [Google Scholar] [CrossRef]
  39. Torgerson, W.S. Theory and Methods of Scaling; John Wiley & Sons: New York, NY, USA, 1958; p. 460. [Google Scholar]
  40. Gower, J.C. Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika 1966, 53, 325–338. [Google Scholar] [CrossRef]
  41. Borg, I.; Groenen, P. Modern Multidimensional Scaling; Springer Science & Business Media: Berlin/Heidelberg, Germany, 1997. [Google Scholar] [CrossRef]
  42. Saeed, N.; Nam, H.; Haq, M.I.U.; Muhammad Saqib, D.B. A Survey on Multidimensional Scaling. ACM Comput. Surv. 2018, 51, 1–25. [Google Scholar] [CrossRef]
  43. Bajusz, D.; Rácz, A.; Héberger, K. Why is Tanimoto index an appropriate choice for fingerprint-based similarity calculations? J. Cheminformatics 2015, 7, 20. [Google Scholar] [CrossRef]
  44. Mayer-Bacon, C.; Meringer, M.; Havel, R.; Aponte, J.C.; Freeland, S. A Closer Look at Non-random Patterns Within Chemistry Space for a Smaller, Earlier Amino Acid Alphabet. J. Mol. Evol. 2022, 90, 307–323. [Google Scholar] [CrossRef]
  45. Bleicher, K.H.; Böhm, H.-J.; Müller, K.; Alanine, A.I. Hit and lead generation: Beyond high-throughput screening. Nat. Rev. Drug Discov. 2003, 2, 369–378. [Google Scholar] [CrossRef]
  46. Wallach, I.; Bernard, D.; Nguyen, K.; Ho, G.; Morrison, A.; Stecula, A.; Rosnik, A.; O’Sullivan, A.M.; Davtyan, A.; Samudio, B.; et al. AI is a viable alternative to high throughput screening: A 318-target study. Sci. Rep. 2024, 14, 7526. [Google Scholar] [CrossRef]
  47. Kun, A.; Papp, B.; Szathmáry, E. Computational identification of obligatorily autocatalytic replicators embedded in metabolic networks. Genome Biol. 2008, 9, R51. [Google Scholar] [CrossRef]
  48. Altschul, S.F. Amino acid substitution matrices from an information theoretic perspective. J. Mol. Biol. 1991, 219, 555–565. [Google Scholar] [CrossRef]
  49. Henikoff, S.; Henikoff, J.G. Amino acid substitution matrices from protein blocks. Proc. Natl. Acad. Sci. USA 1992, 89, 10915–10919. [Google Scholar] [CrossRef]
  50. Sander, T.; Freyss, J.; von Korff, M.; Rufener, C. DataWarrior: An open-source program for chemistry aware data visualization and analysis. J. Chem. Inf. Model. 2015, 55, 460–473. [Google Scholar] [CrossRef]
  51. Tukey, J.W. Exploratory Data Analysis; Addison-Wesley: Reading, MA, USA, 1997; Volume 2, pp. 131–160. [Google Scholar]
  52. DuToit, S.H.; Steyn, A.G.W.; Stumpf, R.H. Graphical Exploratory Data Analysis; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2012. [Google Scholar]
  53. Heckert, N.; Filliben, J.; Croarkin, C.; Hembree, B.; Guthrie, W.; Tobias, P.; Prinz, J. Handbook 151: NIST/SEMATECH e-Handbook of Statistical Methods, NIST Interagency/Internal Report (NISTIR); National Institute of Standards and Technology: Gaithersburg, MD, USA, 2002; Available online: https://www.itl.nist.gov/div898/handbook/prc/section1/prc16.htm (accessed on 20 July 2024).
  54. Hohsaka, T.; Sisido, M. Incorporation of non-natural amino acids into proteins. Curr. Opin. Chem. Biol. 2002, 6, 809–815. [Google Scholar] [CrossRef] [PubMed]
  55. Link, A.J.; Mock, M.L.; Tirrell, D.A. Non-canonical amino acids in protein engineering. Curr. Opin. Biotechnol. 2003, 14, 603–609. [Google Scholar] [CrossRef]
  56. Wang, L.; Schultz, P.G. Expanding the genetic code. Angew. Chem. Int. Ed. 2005, 44, 34–66. [Google Scholar] [CrossRef]
  57. Liu, C.C.; Schultz, P.G. Adding new chemistries to the genetic code. Annu. Rev. Biochem. 2010, 79, 413–444. [Google Scholar] [CrossRef]
  58. Agostini, F.; Völler, J.S.; Koksch, B.; Acevedo-Rocha, C.G.; Kubyshkin, V.; Budisa, N. Biocatalysis with unnatural amino acids: Enzymology meets xenobiology. Angew. Chem. Int. Ed. 2017, 56, 9680–9703. [Google Scholar] [CrossRef]
  59. Bell, E.L.; Finnigan, W.; France, S.P.; Green, A.P.; Hayes, M.A.; Hepworth, L.J.; Lovelock, S.L.; Niikura, H.; Osuna, S.; Romero, E.; et al. Biocatalysis. Nat. Rev. Methods Primers 2021, 1, 46. [Google Scholar] [CrossRef]
Ijms 25 12343 g001
Figure 1. Venn diagram of EAAs, GAAs, and PAAs.
Figure 1. Venn diagram of EAAs, GAAs, and PAAs.
Ijms 25 12343 g001
Ijms 25 12343 g002
Figure 2. Block structure of the similarity matrix. Asterisks represent arbitrary, non-diagonal matrix entries.
Figure 2. Block structure of the similarity matrix. Asterisks represent arbitrary, non-diagonal matrix entries.
Ijms 25 12343 g002
Ijms 25 12343 g003
Figure 3. Cumulative relative frequencies of Tanimoto coefficients for different parts of the similarity matrix. For more explanation, see Section 2.2.
Figure 3. Cumulative relative frequencies of Tanimoto coefficients for different parts of the similarity matrix. For more explanation, see Section 2.2.
Ijms 25 12343 g003
Ijms 25 12343 g004
Figure 4. Structures of the most similar XAAs for each CAA.
Figure 4. Structures of the most similar XAAs for each CAA.
Ijms 25 12343 g004
Ijms 25 12343 g005
Figure 5. Two-dimensional projection of the AA library using Tanimoto distances and classical MDS.
Figure 5. Two-dimensional projection of the AA library using Tanimoto distances and classical MDS.
Ijms 25 12343 g005
Ijms 25 12343 g006
Figure 6. Workflow of the similarity study. For explanation of the abbreviations, see Section 3 and Table 2.
Figure 6. Workflow of the similarity study. For explanation of the abbreviations, see Section 3 and Table 2.
Ijms 25 12343 g006
Table 1. Abbreviations, short descriptions, references or definition, and sizes of the sets of amino acids (AAs) used in this study.
Table 1. Abbreviations, short descriptions, references or definition, and sizes of the sets of amino acids (AAs) used in this study.
AbbreviationDescriptionReference/DefinitionSize
CAAsCanonical AAs (standard code)e.g., Cleaves (2010) [5]20
EAAsExperimentally evaluated AAsJosephson et al. (2005) [19]
Hartman et al. (2007) [20]		105
GAAs(computationally) Generated AAsIlardo et al. (2015) [13]1913
PAAsPurchasable AAswww.emolecules.com9518
AAAsAll AAsEAA ∪ GAA ∪ PAA11,302
XAAsXeno AAsAAA\CAA11,282
Table 2. Statistical parameters of TCs for CAAs, CAA × XAA, XAAs, and AAAs.
Table 2. Statistical parameters of TCs for CAAs, CAA × XAA, XAAs, and AAAs.
ParameterCAA PairsCAA × XAAXAA PairsAAA Pairs
Number of TCs190225,64063,636,12163,861,951
Minimum0.111110.065790.06173 0.06173
First quartile0.257670.137930.125000.12500
Median0.321290.176470.153850.15385
Third quartile0.388440.226420.191180.19149
90% quantile0.463210.281250.232140.23214
99% quantile0.613890.400000.333330.33333
Maximum0.650000.750001.000001.00000
Mean0.328030.189640.163690.16379
Std. deviation0.110540.068600.053010.05309
Table 3. Tanimoto coefficients for CAA. Table cells are colored by a red–yellow–green scale, with red indicating low TC, and green high TC.
Table 3. Tanimoto coefficients for CAA. Table cells are colored by a red–yellow–green scale, with red indicating low TC, and green high TC.
GASPVTCILNDQKEMHFRYW
G1.0000.3330.3500.2080.2860.2730.3330.2800.2920.3180.3810.2590.2860.3200.2410.2420.2060.1940.2000.152
A0.3331.0000.4500.1920.6110.5790.4290.4780.5000.4090.4090.3330.3100.3460.3570.2650.2650.2500.2570.196
S0.3500.4501.0000.1720.3910.3750.5710.3700.5000.5450.5450.4440.4140.4620.4140.3530.3530.3330.3430.261
P0.2080.1920.1721.0000.1720.1670.1670.1470.1520.1610.1610.1390.1320.1430.1320.1430.1160.1110.1140.111
V0.2860.6110.3910.1721.0000.6500.3750.5420.4400.3600.3600.3000.2810.3100.3230.2430.2430.2310.2370.184
T0.2730.5790.3750.1670.6501.0000.3600.5200.4230.3460.3460.2900.2730.3000.3130.2370.2370.2250.2310.180
C0.3330.4290.5710.1670.3750.3601.0000.3570.4800.5220.5220.4290.4000.4440.4000.3820.3430.3240.3330.255
I0.2800.4780.3700.1470.5420.5200.3571.0000.4140.3450.3450.2940.2780.3030.3140.2750.2750.2330.2680.212
L0.2920.5000.5000.1520.4400.4230.4800.4141.0000.4620.4620.3870.3640.4000.4520.3160.3160.3000.3080.240
N0.3180.4090.5450.1610.3600.3460.5220.3450.4621.0000.6360.5190.3870.4290.3870.3330.3330.3510.3240.250
D0.3810.4090.5450.1610.3600.3460.5220.3450.4620.6361.0000.4140.3870.4810.3870.3710.3710.3160.3610.277
Q0.2590.3330.4440.1390.3000.2900.4290.2940.3870.5190.4141.0000.4120.6070.4120.2930.2930.3750.2860.226
K0.2860.3100.4140.1320.2810.2730.4000.2780.3640.3870.3870.4121.0000.4240.3890.2790.3100.4250.3020.241
E0.3200.3460.4620.1430.3100.3000.4440.3030.4000.4290.4810.6070.4241.0000.4240.3000.3000.3500.2930.231
M0.2410.3570.4140.1320.3230.3130.4000.3140.4520.3870.3870.4120.3890.4241.0000.2790.2790.3260.2730.218
H0.2420.2650.3530.1430.2430.2370.3820.2750.3160.3330.3710.2930.2790.3000.2791.0000.3950.2650.3860.333
F0.2060.2650.3530.1160.2430.2370.3430.2750.3160.3330.3710.2930.3100.3000.2790.3951.0000.2920.5640.385
R0.1940.2500.3330.1110.2310.2250.3240.2330.3000.3510.3160.3750.4250.3500.3260.2650.2921.0000.2860.213
Y0.2000.2570.3430.1140.2370.2310.3330.2680.3080.3240.3610.2860.3020.2930.2730.3860.5640.2861.0000.327
W0.1520.1960.2610.1110.1840.1800.2550.2120.2400.2500.2770.2260.2410.2310.2180.3330.3850.2130.3271.000
Table 4. Tanimoto coefficients and set memberships of the most similar amino acids for each CAA.
Table 4. Tanimoto coefficients and set memberships of the most similar amino acids for each CAA.
CAATCIDEAAGAAPAACAATCIDEAAGAAPAA
G0.450G1 x N0.636Dxxx
A0.611A1 xx 0.583N2 xx
0.611V xxD0.636Nxxx
S0.600S1 xx 0.609D2x
P0.750P1x xQ0.607Exxx
0.680P2 x 0.581Q2 x
0.654P3 xK0.714K1 xx
V0.700V1 x E0.607Qxxx
0.684A1 xx 0.586E2 x
0.667V3.1 xM0.606M1 x
0.667V3.2 xH0.590H1 x
T0.714T1 x F0.610F1 x
0.682T2.1 xR0.737R1 x
0.682T2.2 x 0.676R2 x
C0.619C1 xx 0.667R3 x
I0.667I1 x 0.658R4 x
L0.643L1 xY0.649Y1 x
W0.623W1x
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Meringer, M.; Casanola-Martin, G.M.; Rasulev, B.; Cleaves, H.J., II. Similarity Analysis of Computer-Generated and Commercial Libraries for Targeted Biocompatible Coded Amino Acid Replacement. Int. J. Mol. Sci. 2024, 25, 12343. https://doi.org/10.3390/ijms252212343

AMA Style

Meringer M, Casanola-Martin GM, Rasulev B, Cleaves HJ II. Similarity Analysis of Computer-Generated and Commercial Libraries for Targeted Biocompatible Coded Amino Acid Replacement. International Journal of Molecular Sciences. 2024; 25(22):12343. https://doi.org/10.3390/ijms252212343

Chicago/Turabian Style

Meringer, Markus, Gerardo M. Casanola-Martin, Bakhtiyor Rasulev, and H. James Cleaves, II. 2024. "Similarity Analysis of Computer-Generated and Commercial Libraries for Targeted Biocompatible Coded Amino Acid Replacement" International Journal of Molecular Sciences 25, no. 22: 12343. https://doi.org/10.3390/ijms252212343

APA Style

Meringer, M., Casanola-Martin, G. M., Rasulev, B., & Cleaves, H. J., II. (2024). Similarity Analysis of Computer-Generated and Commercial Libraries for Targeted Biocompatible Coded Amino Acid Replacement. International Journal of Molecular Sciences, 25(22), 12343. https://doi.org/10.3390/ijms252212343

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop