ActSeek: Fast and accurate search algorithm of active sites in Alphafold database

ActSeek: Fast and accurate search algorithm of active sites in Alphafold database | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results ActSeek: Fast and accurate search algorithm of active sites in Alphafold database View ORCID Profile Sandra Castillo , View ORCID Profile O. H. Samuli Ollila doi: https://doi.org/10.1101/2025.02.11.637678 Sandra Castillo 1 VTT Technical Research Centre of Finland Ltd , Espoo, Finland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sandra Castillo For correspondence: sandra.castillo{at}vtt.fi O. H. Samuli Ollila 1 VTT Technical Research Centre of Finland Ltd , Espoo, Finland 2 Institute of Biotechnology, University of Helsinki , Finland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for O. H. Samuli Ollila Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract Finding proteins with specific functions by mining modern databases can potentially lead to substantial advancements in wide range of fields, from medicine and biotechnology to material science. Currently available algorithms enable mining of proteins based on their sequence or structure. However, activities of many proteins, such as enzymes and drug targets, are dictated by active site residues and their surroundings rather than the overall structure or sequence of a protein. Here we present ActSeek – a computer vision-inspired fast program – that searches structural databases for proteins with active sites similar to the seed protein. ActSeek is implemented to mine proteins with desired active site environments from the Alphafold database. The potential of ActSeek to find innovative solutions to the world’s most pressing challenges is demonstrated by finding enzymes that may be used to produce biodegradable plastics or degrade plastics, as well as potential off-targets for common drug molecules. Introduction Finding proteins with desired functions by mining databases is becoming increasingly attractive due to the growing amount data and emerging computational tools. In addition to Uniprot database ( 1 ) containing protein sequences, also in silico predicted protein structures are available for mining ( 2 , 3 ). These databases open new avenues for searching proteins based also on structural similarity, in addition to traditional sequence based approaches using BLAST or similar algorithms ( 4 ). Such algorithms are available, but their computational efficiency is often a limiting factor for large scale structure based searches ( 5 – 8 ). A significant improvement in performance was presented in the recent FoldSeek algorithm ( 7 ), which enabled structure-based searches over a large number of structures available in databases. However, the activities of some relevant proteins, such as enzymes and drug targets, is dictated by several residues in the proximity of active sites rather than the overall structure of a protein. Therefore, searches based on full structure may result in proteins whose overall structure resembles the seed protein, but have critical differences at the active site region. On the other hand, proteins with similar active site to the seed protein but some irrelevant structural differences may not be found. Here we present ActSeek, a program that searches structural databases for proteins with similar active site to the seed protein. Our fast algorithm, inspired by computer vision ( 9 , 10 ), enables searches over large databases based on the coordinates of specific amino acids. Here we have implemented ActSeek to search over the Alphafold database ( 3 ). Finding proteins with desired active sites and functions can benefit wide range of fields, from medicine and biotechnology to material science. Here we demonstrate the potential of ActSeek to find (i) proteins that are able to produce biodegradable plastics with enhanced properties (polyhydroxyalkanoate (PHA) synthases) ( 11 , 12 ), (ii) new proteins to degrade poly(ethylene terephthalate) (PET) based plastics (depolymerases PETase and MHETase) bearing potential in recycling applications ( 13 ), and (iii) potential off-targets for common drugs. These examples demonstrate the potential of protein mining based on active site structures in providing innovative solutions to the world’s most pressing challenges. While other methods can find structural motifs with high speed as ActSeek does ( 14 – 16 ), they are limited to smaller databases such as the RCSB PDB ( 17 ) or BRENDA ( 18 ). This limitation makes it difficult to search larger or customized protein databases. ActSeek, on the other hand, searches in the AlphaFold database ( 3 ), which contains over 200 million proteins, or any protein set the user provides. Results and Discussion A. ActSeek algorithm In contrast to BLAST ( 4 ) and Fold-Seek ( 7 ) that are based on overall sequence or structure of proteins, ActSeek finds proteins with similar local structure to the active site of a seed protein ( Fig. 1A ). Here we implement the search from the Alphafold database ( 3 ), yet the algorithm can be applied to other databases with 3D structures of proteins under interest. Download figure Open in new tab Fig. 1. A Structure of an example seed protein with active site amino acids marked with red. B Query proteins are taken from a data (alphafold database ( 3 ) in this work). C Optimal rotational and translational matrices found with SVD and least square fitting ( 10 ). D Best targets selected based on local structure. To initiate the ActSeek search, user needs to first provide the structure of the seed protein and define the relevant amino acids for the active site and its surroundings ( Fig. 1A ). Alternatively, the user can provide the coordinates of the alpha and beta carbons of the relevant amino acids. Three most relevant amino acids defined by the user, corresponding for example catalytic or binding site, will be used to compute the optimal rotational and translational matrices for structural alignment of the seed and query proteins using singular value decomposition (SVD) and least square fitting ( 10 ) ( Fig. 1C ). After euclidean transformation using these matrices, query proteins found from a given database are ranked based on distances between amino acids in the seed and the query structures. These distances are calculated using binding site amino acids together with additional ones defined by the user, and their neighboring residues. As a results, ActSeek provides a ranked list of natural enzymes with similar 3D arrangement of the active site residues to the seed ( Fig. 1C ). To maximize the speed of the algorithm and enable searches over large datasets, query sequences are pre-screened before proceeding to computationally expensive comparison of 3D arrangements between seed and query proteins. This is done by checking that a query structure contains the three active site amino acids with mutual distances similar to the seed. B. Searching new enzymes To demonstrate the utility of ActSeek for finding new enzymes, we searched proteins that contain active sites characteristic for PHA synthases ( 19 ), and two different PET depolymerases PETase and MHETase ( 20 ). PHA synthases are used to produce biodegradable plastics ( 11 , 12 ), while PETase and MHETase degrade widely used plastics bearing potential in recycling applications ( 13 , 21 ). ActSeek results for these enzymes are compared with the sequence based BLAST ( 4 ) and structure based FoldSeek ( 7 ) in Fig. 2 a . Download figure Open in new tab Fig. 2. A Results for ActSeek searches over proteins in all organisms using active sites of a PHA synthase, PETase and MHETase as seeds. B Results for ActSeek searches over human proteins with relevant binding pocket amino acids for eriotinib, sorafenib and beta blockers as seeds. For all these enzymes, ActSeek finds significant amounts of proteins that are not recognized by BLAST or FoldSeek searches (11964 for PHA synthase, 11115 for PETase and 216 for MHETase). This means that large number of enzymes containing the relevant active site, thereby potentially having the desired functions, are not found by BLAST or FoldSeek searches. On the other hand, these methods find many proteins with similar sequences or overall structures to the seed, but that do not contain the active site required for the desired reaction. These results demonstrate that ActSeek can 1) find new proteins with the desired active site that are not found by searches based on sequences or overall structures, and 2) filter out proteins that do not contain the relevant active sites from BLAST or FoldSeek results. The new natural enzymes found by ActSeek can bear novel properties that are potentially useful in wide range of applications from specificity for relevant substrates to functionality in harsh conditions. Furthermore, our results demonstrate the potential of Act-Seek to improve functional annotation of proteins. For example, 31585 proteins are currently annotated as PHA synthases in the Uniprot database ( 1 ). However, 9080 of these do not contain the relevant active site according to the ActSeek. On the other hand, ActSeek finds 4875 proteins with the relevant active site for PHA synthesis, but that are not annotated for this functions in Uniprot database. Based on these results, ActSeek can provide new annotations for protein functions based on active sites, as well as filter incorrect annotations from databases. C. Searching off-target proteins for drugs Beside enzymes, drug binding to query proteins is another example where local binding region is more relevant than the over-all structure or sequence. Here we demonstrate that Act-Seek can find potential off-target proteins for drug molecules. To this end, we searched all human proteins from Alphafold database with binding pockets for two anti-cancer drugs, Erlotinib ( 22 ) and sorafenib ( 23 ), and for beta-blockers used to treat various cardiovascular diseases such as hypertension, cardiac arrhythmia, or myocardial infarction and also alleviate severe anxiety ( 24 ). Erlotinib and sorafenib inhibit epidermal growth factors and protein kinases, respectively, and beta-blockers block receptor sites of adrenergic receptors. ActSeek results using the main targets of these drugs as seeds (Uniprot P00533, P15056, P07550, respectively) are compared with the sequence based BLAST ( 4 ) and structure based FoldSeek ( 7 ) in Fig. 2 b . ActSeek finds relatively few proteins with suitable active sites for binding of drugs studied here (107 for Erlotinib, 12 for Sorafenib and 13 for beta blocker). Majority of ActSeek results are found also by Blast search based on sequences, but Blast finds also hundreds of proteins that do not contain the relevant binding site according to ActSeek, therefore being unlikely targets for the given drug ( Fig. 2b ). Structure based FoldSeek search finds hundreds of proteins with similar structure to the seed, but only few of them (17 for Erlotinib and one for beta-blocker off-targets) contain the relevant binding site, compromising its suitability for finding drug off-targets. For Erlotinib, ActSeek results contain two proteins that were not identified with sequence based Blast search ( Fig. 2 ). These two proteins have less than 30% sequence similarity with the seed protein and only partial structural similarity, but they have relevant amino acids in the exact locations as in the seed protein ( Fig. 3 ), which explains why they were found by ActSeek but not by Blast or FoldSeek. Previously reported off-targets for Erlotinib that may contribute to the common side-effects ( 25 , 26 ) were identified by both Act-Seek and Blast, but not by FoldSeek. Download figure Open in new tab Fig. 3. Potential off-target for Erlotinib, tyrosine-protein kinase Srms (Uniprot: Q9H3Y6) (pink) superposed to the seed protein epidermal growth factor receptor (Uniprot: P00533) (light green). This protein was found by ActSeek but not by Blast or FoldSeek. The amino acids used for the search of Erlotinib off-targets are marked as blue (718Leu, 790Thr and 855Asp). The corresponding amino acids in the tyrosine-protein kinase Srms are marked as red (236Leu, 302Thr and 368Asp). For Sorafenib, we identified 11 potential off-targets. These include the retinal guanylyl cyclase2 (Uniprot P51841) found also by Blast, which may explain one of the reported drug’s side effects related to the retinal tear ( 27 ). Additionally, two of the identified drug targets are annotated as atrial natriuretic peptide receptor 1 and 2 (Uniprot P16066 and P20594) and were found only by ActSeek. Alignment of atrial natriuretic peptide receptor 1 with the main Sorafenib target, B-Raf kinase (Uniprot P15056), with Sorafenib docked is shown in Fig. 4 . Inhibiting these receptors may cause high blood pressure and heart-related problems, which are also reported side effects of the drug ( 28 ). However, none of these receptors have been previously reported as off-targets of the drug. Also, previously reported off-targets for Sorafenib, VEGFR 1, 2 and 3 ( 29 ), are among the ActSeek results. Download figure Open in new tab Fig. 4. Potential off-target for Sorafenib, atrial natriuretic peptide receptor 1 (Uniprot P16066, beige), aligned with the main target, B-Raf kinase (Uniprot P15056, light green). Amino acids involved in the binding are marked as red in the B-Raf kinase and blue in the atrial natriuretic peptide receptor 1. Sorafenib (magenta) has been docked to the B-Raf kinase structure. All 12 potential off-targets for beta-blockers found by Act-Seek are identified also by Blast. Successfully recognized case by Blast with relevant active site, 5-hydroxytryptamine receptor, is superposed with the seed in Fig. 5B . However, the usefulness for sequence based search for this task is compromised by a large number of false-positive results. For example, Trace amine-associated receptor 2 (Uniprot Q9P1P5) is ranked high in the Blast results (E-value: 4.2 × 10 −49 ) and has similar overall structure to the seed ( Fig. 5A ), but does not contain 290Phe and 312Asn amino acids which are involved in the binding of the beta-blocker drug ( Fig. 5C ). Act-Seek results for beta-blockers include different types of beta adrenergic (1,2 and 3), serotonine receptors and dopaminergic receptors, among others. Interaction of beta-blockers with dopaminergic receptors has been previously reported ( 30 ), yet specific receptors found by Actseek are potentially novel off-targets for beta-blockers. Download figure Open in new tab Fig. 5. Comparison of results from ActSeek and Blast for beta-blocker off-targets. A Superposition of the alphafold structures for the seed (beta2 adrenergic receptor (P07550), (pink), for a protein found by both ActSeek and Blast (5-hydroxytryptamine receptor 1B (P08908), (beige), and a protein found only by Blast but not ActSeek, Trace amine-associated receptor 2 (Q96RI9), (green). B Super-position of active site amino acids from the seed (red), and 5-hydroxytryptamine receptor 1B (blue) found by both Actseek and Blast. C Superposition of active site amino acids from the seed (red) and the Trace amine-associated receptor 2 (Q96RI9)(blue), which was found by Blast but not by Actseek. Conclusions We present here the ActSeek algorithm to rapidly search proteins with relevant active sites from alphafold database based on seed protein coordinates. Advantages of ActSeek to commonly used protein mining tools based on sequences (BLAST) and structures (FoldSeek) are demonstrated for finding new enzymes and drug off-targets. Our results demonstrate that ActSeek can complement searches based on sequence or overall structure by finding proteins containing the relevant active site that are not recognized based on sequence or structure. On the other hand, ActSeek can refine searches by filtering out results with similar sequence or structure, but not having the relevant active site. These features increase the probability to find new proteins with desired functions, particularly when structural details of a local active site are crucial for function. This helps to focus further screening efforts to most probable candidates. Finding proteins with matching active site 3D conformations is particularly useful for proteins whose functions are based on specific binding certain substances, such as enzymes and drug targets. For example, ActSeek can find natural proteins from databases with similar active sites to the seed enzyme, which can be then further screened for relevant novel functionalities, such as specificity for new substrates. Expressibility and biocompatibility of such natural proteins are expected to be on average better than for de novo designed proteins. Another example on practical benefit is finding potential off-targets for drugs by searching human proteins with similar binding sites to known targets. Sequence based search returns many false positive results for such search, but Act-Seek can filter out results that lack the relevant active sites, thereby substantially reducing the resources required for further screening. We foresee wide range of applications for ActSeek in understanding and engineering biological systems from classification of protein functionalities for functional annotation to protein engineering, enzyme design and drug development. Supplementary Note 1: Material and Methods A. Defining the relevant amino acids in the active site For a start, the user needs to define the binding site in the seed protein by providing cartesian coordinates or positions of the three most important amino acids. These amino acids will be used to align the query structure with the seed. User can also define additional amino acids that are relevant for the active site, which will be used for ranking when selecting the best query protein. Also neighboring amino acids in the sequence (four by default, but can be reset by the user) to the selected amino acids (binding site and additional ones) will be used for the ranking. Coordinates of C α and C β atoms in the amino acids will be used in the searches. User can define whether amino acid types in the local structure of the query need to exactly correspond the seed ones, or define a list of similar amino acids that are accepted at each position. B. Pre-screening ActSeek algorithm first downloads the query structure from the database and checks if the three main amino acids defined by the user are present in the sequence. Before comparing local 3D configurations with the seed, ActSeek filters potential matches by comparing interamino acid distances. To this end, distances between the amino acids from all possible sequential combinations are calculated. Only if all inter-amino acid distances deviate less than 3Å from the values in the seed structure, the structure with the given mapping will be considered further. Notably, several possible mappings may be found for a single structure. If the number of possible combinations exceeds 2000 (can be reset by the user), this amount of combinations are randomly selected for testing. C. Finding the optimal rotation and translation matrices Local 3D structures that satisfy the condition for binding site inter-amino acid distances in the pre-screening step are then further compared with the seed. This is done by finding the optimal rotation and translation matrices to over-lay the binding site amino acid configuration with the seed using singular value decomposition (SVD) and least square fitting ( 9 ). D. Selecting the best query proteins After performing euclidean transformations with the optimal matrices, similarities of active sites between query and seed proteins are measured by calculating distances between relevant amino acids. To this end, we first calculate the average distances for selected amino acids (binding site and additional ones) between a query structure and the seed. Additional amino acids are counted only if corresponding type is found from the query protein (exact match or correspondence defined by the user). In addition, the same distances are calculated for neighboring amino acids (set by user or 4 by default) around the each relevant amino acids. Distances between neighboring amino acids are used regardless of the types of these amino acids in the query and seed. The score for each possible mapping of amino acids in the given structure is then calculated as an average of the distances of mapped amino acids divided by the number of amino acids over which the average is taken (to penalize situations where additional amino acids defined by the user are not found from the query). The combination of mapped amino acids with lowest score is then selected for the given structure. Finally, the given structure is selected if its lowest score was less than user defined value (default 1 Å). E. Other calculations. ActSeek calculates additional measures to assess the structural similarity between seed and query proteins, which can help filter the results further Protein alignment percentage: The alpha carbons of the query structure are aligned with those of the seed protein by optimally pairing them using dynamic programming ( 31 ). Only pairs with the distance below 2 Å are considered and used to calculate the percentage of aligned amino acids in the query structure. This value is given as a “Structural mapping percentage” in the result file. Local protein alignment: We calculate a weighted average of the number of 5, 10, 15, and 20 consecutive amino acid pairs that are aligned in both structures. This average is then normalized with the value from perfectly aligned structures, resulting values between 0 (local similarities not found) and 1 (structures fully aligned). This value is given by ActSeek as “Structural local similarity” in the result file. This measure indicates similarities of substructures that are aligned with the seed protein, without considering the rest of the protein that might differ significantly. Cavity Comparison: To compare active site cavities between query and seed proteins, cavities are first recognized using the pyKVFinder ( 32 ) algorithm. Cavities between seed and query are then compared based on above described protein alignment. We also provide the percentage of amino acids aligned in the cavity compared to the total amino acids. The result file provides these measures as ‘Cavity distances’, ‘Cavity mapping (case:seed)’, and ‘Cavity mapping percentage’. This feature is optional because it reduces the program’s performance during extensive searches. F. Implementation and running The algorithm supports multiprocessing and it is designed for computer clusters, though it can run in a single computer for smaller searches. Each running node in the cluster with 24 CPUs would process 10000 proteins per 2-10 minutes. with timing dependent on the average number of possible mappings to test. ActSeek is written using python. The code can be found in https://github.com/vttresearch/ActSeek G. Blast searches and FoldSeek searches We used Uniprot Knowledgebase TrEMBLE ( 1 ), containing around 253 millions of sequences as a database for the Blast search. The Evalue threshold was set to 10. We used Blastp algorithm ( 4 ) for the protein search. For the FoldSeek ( 7 ) search we used the Alphafold/UniProt50 database ( 3 ) and run it with the default parameters H. ActSeek search examples PHA synthases We used the PHA synthases Class I from Chromobacterium violaceum (uniprot: Q9ZHI2) as the seed, and its three catalytic amino acids 291Cys, 447Asp, and 477His ( 33 ) as the sites for the ActSeek search. Additional amino acids defined for the search were 393Trp, and 477Gly. Default values were used for other ActSeek parameters. The search was performed through Alphafold database( 3 ) for proteins that are annotated with αβ -hydrolase fold in the Uniprot database ( 1 ) (total 2567710 enzymes) because all PHA synthases are expected to have such fold. We filtered the data based on local protein alignment score because we did not anticipate significant structural divergence in this case. When the mapping included only the three main catalytic amino acids (additional amino acids not found), we set the threshold for local protein alignment score to 0.4. When additional amino acids were included in the mapping, we lowered the threshold to 0.1. After filtering, we also removed any proteins that had been deleted from UniProt during the course of our study. PETase We used poly(ethylene terephthalate) (PET) depolymerase from Piscinibacter sakaiensis ( Ideonella sakaiensis ) as the PETase seed (Uniprot: A0A0K8P6T7), and its three catalytic amino acids 160Ser, 206Asp, and 237His ( 34 ) as the sites for the ActSeek search. Additional amino acids defined for the search were 87Tyr and 185Trp. Also phenylalanine was accepted for position 87 and tyrosine for position 185. Five neighboring amino acids around each each selected amino acids were used for the ranking (instead of default value of four). The search was performed for proteins annotated with αβ -hydrolase fold in the Uniprot database ( 1 ) (total 2567710 enzymes) as PETases are know to have the αβ -hydrolase fold ( 34 ). MHETase We used protein that catalyzes the hydrolysis of mono(2-hydroxyethyl) terephthalate (MHET) as the MHETase seed (uniprot: A0A0K8P8E7), and its amino acids 225Ser, 492Asp, and 528His as the sites for the search ( 35 ). Additional amino acids for the search were 224Cys and 529Cys. Default values were used for other ActSeek parameters. The search was performed in proteins annotated with αβ -hydrolase fold in the Uniprot database ( 1 ) (total 2567710 enzymes) as MHETases are know to have the αβ -hydrolase fold. Erlotinib targets We performed a search using the target of the drug Erlotinib as a seed (Epidermal growth factor receptor, Uniprot P00533). For the search, we used important amino acids for the binding of the drug, such as 718Leu, 790Thr, and 855Asp ( 36 ) without additional ones. Also methionine or isoleucine were accepted at position 790. After the search, we filtered the results by removing hits with the local protein alignment score lower than 0.4. We searched in all the human proteins listed in Uniprot (20,417 proteins). Sorafenib targets We performed a search using the Sorafenib target B-Raf kinase as a seed (Uniprot P15056), and its amino acids 529Thr, 532Cys and 536Ser as the sites used for the search without any additional amino acid. Also valine was accepted at position 529 and asparagine at position 536. We searched in all the human proteins listed in Uniprot (20,417 proteins). Beta-blocker targets We used the native receptor of these drugs, the beta2 adrenergic receptor (Uniprot id: P07550) as the seed. The main amino acids used for the search were 113Asp, 290Phe and 312Asn and additional ones were 114Val and 289Phe ( 37 ), which are known to be involved in the drug binding. We allowed amino acids of the same type (i.e., uncharged polar, charged, hydrophobic, and aromatic) to be interchangeable during the search. We searched in all the human proteins listed in Uniprot (20,417 proteins). We are making the results of ActSeek, Blast and FoldSeek available in the supplementary material. Supplementary Note 2: Acknowledgements We acknowledge CSC – IT Center for Science for computational resource and the Jane and Aatos Erkko Foundation (JAES) under project 220048 (Virtual laboratory for Biodesign, JAES-BIODESIGN) for their important support that allowed the realization of this work. OHSO acknowledges Research Council of Finland for financial support (358505, 356568). Additionally, we would like to acknowledge Antonio L. Rodríguez and Alejandro Revuelta for their help and advise related to the computer vision algorithms and code optimization, and Gopal Peddinti for his help running FoldSeek. Bibliography 1. ↵ The UniProt Consortium . UniProt: the Universal Protein Knowledgebase in 2023 . Nucleic Acids Research , 51 ( D1 ): D523 – D531 , 11 2022 . ISSN 0305-1048 . doi: 10.1093/nar/gkac1052 . OpenUrl CrossRef PubMed 2. ↵ John Jumper , Richard Evans , Alexander Pritzel , Tim Green , Michael Figurnov , Olaf Ronneberger , Kathryn Tunyasuvunakool , Russ Bates , Augustin Žídek , Anna Potapenko , Alex Bridgland , Clemens Meyer , Simon A. A. Kohl , Andrew J. Ballard , Andrew Cowie , Bernardino Romera-Paredes , Stanislav Nikolov , Rishub Jain , Jonas Adler , Trevor Back , Stig Petersen , David Reiman , Ellen Clancy , Michal Zielinski , Martin Steinegger , Michalina Pacholska , Tamas Berghammer , Sebastian Bodenstein , David Silver , Oriol Vinyals , Andrew W. Senior , Koray Kavukcuoglu , Pushmeet Kohli , and Demis Hassabis . Highly accurate protein structure prediction with AlphaFold . Nature , 596 ( 7873 ): 583 – 589 , August 2021 . ISSN 1476-4687 . doi: 10.1038/s41586-021-03819-2 . Number: 7873 Publisher: Nature Publishing Group . OpenUrl CrossRef PubMed 3. ↵ Mihaly Varadi , Damian Bertoni , Paulyna Magana , Urmila Paramval , Ivanna Pidruchna , Malarvizhi Radhakrishnan , Maxim Tsenkov , Sreenath Nair , Milot Mirdita , Jingi Yeo , Oleg Kovalevskiy , Kathryn Tunyasuvunakool , Agata Laydon , Augustin Žídek , Hamish Tomlinson , Dhavanthi Hariharan , Josh Abrahamson , Tim Green , John Jumper , Ewan Birney , Martin Steinegger , Demis Hassabis , and Sameer Velankar . AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences . Nucleic Acids Research , 52 ( D1 ): D368 – D375 , January 2024 . ISSN 0305-1048 . doi: 10.1093/nar/gkad1011 . OpenUrl CrossRef 4. ↵ Stephen F. Altschul , Warren Gish , Webb Miller , Eugene W. Myers , and David J. Lipman . Basic local alignment search tool . Journal of Molecular Biology , 215 ( 3 ): 403 – 410 , October 1990 . ISSN 0022-2836 . doi: 10.1016/S0022-2836(05)80360-2 . OpenUrl CrossRef PubMed Web of Science 5. ↵ Pieter Meysman , Cheng Zhou , Boris Cule , Bart Goethals , and Kris Laukens . Mining the entire protein databank for frequent spatially cohesive amino acid patterns . BioData Mining , 8 ( 1 ): 4 , 2015 . ISSN 1756-0381 . doi: 10.1186/s13040-015-0038-4 . OpenUrl CrossRef PubMed 6. Gabriel Cia , Jean Kwasigroch , Basile Stamatopoulos , Marianne Rooman , and Fabrizio Pucci . pyscomotif: discovery of similar 3d structural motifs across proteins . Bioinformatics Advances , 3 ( 1 ): vbad158 , 11 2023 . ISSN 2635-0041 . doi: 10.1093/bioadv/vbad158 . OpenUrl CrossRef 7. ↵ Michel van Kempen , Stephanie S. Kim , Charlotte Tumescheit , Milot Mirdita , Jeongjae Lee , Cameron L. M. Gilchrist , Johannes Söding , and Martin Steinegger . Fast and accurate protein structure search with foldseek . Nature Biotechnology , 42 ( 2 ): 243 – 246 , Feb 2024 . ISSN 1546-1696 . doi: 10.1038/s41587-023-01773-0 . OpenUrl CrossRef PubMed 8. ↵ Stephen K Burley , Rusham Bhatt , Charmi Bhikadiya , Chunxiao Bi , Alison Biester , Pratyoy Biswas , Sebastian Bittrich , Santiago Blaumann , Ronald Brown , Henry Chao , Vivek Reddy Chithari , Paul A Craig , Gregg V Crichlow , Jose M Duarte , Shuchismita Dutta , Zukang Feng , Justin W Flatt , Sutapa Ghosh , David S Goodsell , Rachel Kramer Green , Vladimir Guranovic , Jeremy Henry , Brian P Hudson , Michael Joy , Jason T Kaelber , Igor Khokhriakov , Jhih-Siang Lai , Catherine L Lawson , Yuhe Liang , Douglas Myers-Turnbull , Ezra Peisach , Irina Persikova , Dennis W Piehl , Aditya Pingale , Yana Rose , Jared Sagendorf , Andrej Sali , Joan Segura , Monica Sekharan , Chenghua Shao , James Smith , Michael Trumbull , Brinda Vallat , Maria Voigt , Ben Webb , Shamara Whetstone , Amy Wu-Wu , Tongji Xing , Jasmine Y Young , Arthur Zalevsky , and Christine Zardecki . Updated resources for exploring experimentally-determined pdb structures and computed structure models at the rcsb protein data bank . Nucleic Acids Research , 53 ( D1 ): D564 – D574 , 11 2024 . ISSN 1362-4962 . doi: 10.1093/nar/gkae1091 . OpenUrl CrossRef 9. ↵ K. S. Arun , T. S. Huang , and S. D. Blostein . Least-squares fitting of two 3-d point sets . IEEE Transactions on Pattern Analysis and Machine Intelligence , PAMI- 9 ( 5 ): 698 – 700 , 1987 . doi: 10.1109/TPAMI.1987.4767965 . OpenUrl CrossRef PubMed Web of Science 10. ↵ Gene H. Golub and Charles F. Van Loan . Matrix Computations - 4th Edition . Johns Hopkins University Press , Philadelphia, PA , 2013 . doi: 10.1137/1.9781421407944 . OpenUrl CrossRef 11. ↵ Palmiro Poltronieri and Prasun Kumar . Polyhydroxyalcanoates (PHAs) in Industrial Applica-tions , pages 1 – 30 . Springer International Publishing, Cham , 2017 . ISBN 978-3-319-48281-1 . doi: 10.1007/978-3-319-48281-1_70-1 . OpenUrl CrossRef 12. ↵ Matheus Silva da Fonseca Diniz , Murilo Moraes Mourão , Luciana Pereira Xavier , and Agenor Valadares Santos . Recent biotechnological applications of polyhydroxyalkanoates (pha) in the biomedical sector—a review . Polymers , 15 ( 22 ), 2023 . ISSN 2073-4360 . doi: 10.3390/polym15224405 . OpenUrl CrossRef 13. ↵ Writtik Maity , Subhasish Maity , Soumen Bera , and Amrita Roy . Emerging Roles of PETase and MHETase in the Biodegradation of Plastic Wastes . Applied Biochemistry and Biotechnology , 193 ( 8 ): 2699 – 2716 , August 2021 . ISSN 1559-0291 . doi: 10.1007/s12010-021-03562-4 . OpenUrl CrossRef PubMed 14. ↵ Georg Steinkellner , Christian C. Gruber , Tea Pavkov-Keller , Alexandra Binter , Kerstin Steiner , Christoph Winkler , Andrzej Lyskowski , Orsolya Schwamberger , Monika Oberer , Helmut Schwab , Kurt Faber , Peter Macheroux , and Karl Gruber . Identification of promiscu-ous ene-reductase activity by mining structural databases using active site constellations . 5 ( 1 ): 4150 . ISSN 2041-1723 . doi: 10.1038/ncomms5150 . Publisher: Nature Publishing Group . OpenUrl CrossRef PubMed 15. Sebastian Bittrich , Stephen K. Burley , and Alexander S. Rose . Real-time structural motif searching in proteins using an inverted index strategy . 16 ( 12 ): e1008502 . ISSN 1553-7358 . doi: 10.1371/journal.pcbi.1008502 . Publisher: Public Library of Science . OpenUrl CrossRef PubMed 16. ↵ Zhenkun Shi , Dehang Wang , Yang Li , Rui Deng , Jiawei Lin , Cui Liu , Haoran Li , Ruoyu Wang , Muqiang Zhao , Zhitao Mao , Qianqian Yuan , Xiaoping Liao , and Hongwu Ma . REME: an integrated platform for reaction enzyme mining and evaluation . 52 : W299 – W305 . ISSN 0305-1048 . doi: 10.1093/nar/gkae405 . OpenUrl CrossRef 17. ↵ Helen M. Berman , John Westbrook , Zukang Feng , Gary Gilliland , T. N. Bhat , Helge Weissig, Ilya N. Shindyalov, and Philip E. Bourne . The protein data bank . 28 ( 1 ): 235 – 242 . ISSN 0305-1048 . doi: 10.1093/nar/28.1.235 . OpenUrl CrossRef PubMed Web of Science 18. ↵ Antje Chang , Lisa Jeske , Sandra Ulbrich , Julia Hofmann , Julia Koblitz , Ida Schomburg , Meina Neumann-Schaal , Dieter Jahn , and Dietmar Schomburg . BRENDA, the ELIXIR core data resource in 2021: new developments and updates . 49 : D498 – D508 . ISSN 0305-1048 . doi: 10.1093/nar/gkaa1025 . OpenUrl CrossRef 19. ↵ Alexander Steinbüchel , El-Mehdi Debzi , Robert H. Marchessault , and Arnulf Timm . Syn-thesis and production of poly(3-hydroxyvaleric acid) homopolyester by Chromobacterium violaceum . Applied Microbiology and Biotechnology , 39 ( 4 ): 443 – 449 , July 1993 . ISSN 1432-0614 . doi: 10.1007/BF00205030 . OpenUrl CrossRef 20. ↵ Shosuke Yoshida , Kazumi Hiraga , Toshihiko Takehana , Ikuo Taniguchi , Hironao Yamaji , Yasuhito Maeda , Kiyotsuna Toyohara , Kenji Miyamoto , Yoshiharu Kimura , and Kohei Oda . A bacterium that degrades and assimilates poly(ethylene terephthalate ). Science , 351 ( 6278 ): 1196 – 1199 , March 2016 . doi: 10.1126/science.aad6359 . Publisher: American Association for the Advancement of Science . OpenUrl Abstract / FREE Full Text 21. ↵ Hogyun Seo , Hwaseok Hong , Jiyoung Park , Seul Hoo Lee , Dongwoo Ki , Aejin Ryu , Hye-Young Sagong , and Kyung-Jin Kim . Landscape profiling of PET depolymerases using a natural sequence cluster framework . 387 ( 6729 ): eadp5637 . doi: 10.1126/science.adp5637 . Publisher: American Association for the Advancement of Science . OpenUrl CrossRef 22. ↵ M. A. Bareschino , C. Schettino , T. Troiani , E. Martinelli , F. Morgillo , and F. Ciardiello . Erlotinib in cancer treatment . Annals of Oncology , 18 : vi35 – vi41 , June 2007 . ISSN 0923-7534 , 1569-8041. doi: 10.1093/annonc/mdm222 . Publisher: Elsevier . OpenUrl CrossRef PubMed 23. ↵ Mariko Yoshida , Taro Yamashita , Hikari Okada , Naoki Oishi , Kouki Nio , Takehiro Hayashi , Yoshimoto Nomura , Tomoyuki Hayashi , Yoshiro Asahina , Mika Ohwada , Ha-jime Sunagozaka , Hajime Takatori , Federico Colombo , Laura Porretti , Masao Honda , and Shuichi Kaneko . Sorafenib suppresses extrahepatic metastasis de novo in hepatocellular carcinoma through inhibition of mesenchymal cancer stem cells characterized by the ex-pression of CD90 . Scientific Reports , 7 ( 1 ): 11292 , September 2017 . ISSN 2045-2322 . doi: 10.1038/s41598-017-11848-z . Publisher: Nature Publishing Group . OpenUrl CrossRef PubMed 24. ↵ Luc Poirier and Sheldon W. Tobe . Contemporary use of β-blockers: Clinical relevance of subclassification . Canadian Journal of Cardiology , 30 ( 5 , Supplement): S9 – S15 , 2014 . ISSN 0828-282X . doi: 10.1016/j.cjca.2013.12.001 . β-Adrenergic Receptor Blockers in Hypertension. OpenUrl CrossRef PubMed 25. ↵ Simone Boehrer , Lionel Adès , Thorsten Braun , Lorenzo Galluzzi , Jennifer Grosjean , Claire Fabre , Génèviève Le Roux , Claude Gardin , Antoine Martin , Stéphane de Botton , Pierre Fenaux , and Guido Kroemer . Erlotinib exhibits antineoplastic off-target effects in AML and MDS: a preclinical study . Blood , 111 ( 4 ): 2170 – 2180 , February 2008 . ISSN 0006-4971 . doi: 10.1182/blood-2007-07-100362 . OpenUrl Abstract / FREE Full Text 26. ↵ Naoko Yamamoto , Masashi Honma , and Hiroshi Suzuki . Off-Target Serine/Threonine Ki-nase 10 Inhibition by Erlotinib Enhances Lymphocytic Activity Leading to Severe Skin Dis-orders . Molecular Pharmacology , 80 ( 3 ): 466 – 475 , September 2011 . ISSN 0026895X . doi: 10.1124/mol.110.070862 . OpenUrl Abstract / FREE Full Text 27. ↵ Kelly M. Gaertner , Stephen H. Caldwell , and Osama E. Rahma . A Case of Retinal Tear Associated with Use of Sorafenib . Frontiers in Oncology , 4 , July 2014 . ISSN 2234-943X . doi: 10.3389/fonc.2014.00196 . Publisher: Frontiers . OpenUrl CrossRef 28. ↵ Michael L. Maitland , Kristen E. Kasza , Theodore Karrison , Kristin Moshier , Laura Sit , Henry R. Black , Samir D. Undevia , Walter M. Stadler , William J. Elliott , and Mark J. Ratain . Ambulatory monitoring detects sorafenib-induced blood pressure elevations on the first day of treatment . 15 ( 19 ): 6250 – 6257 . ISSN 1078-0432 . doi: 10.1158/1078-0432.CCR-09-0058 . OpenUrl Abstract / FREE Full Text 29. ↵ Scott Wilhelm , Christopher Carter , Mark Lynch , Timothy Lowinger , Jacques Dumas , Roger A. Smith , Brian Schwartz , Ronit Simantov , and Susan Kelley . Discovery and development of sorafenib: a multikinase inhibitor for treating cancer . 5 ( 10 ): 835 – 844 . ISSN 1474-1784 . doi: 10.1038/nrd2130 . Publisher: Nature Publishing Group . OpenUrl CrossRef PubMed Web of Science 30. ↵ Shlomo Yehuda Yehuda Ben-uriah and Ralph L. Carasso . Physiological evidence (hypother-mia) for central effects of beta-blocker agents . International Journal of Neuroscience , 14 ( 1-2 ): 41 – 46 , 1981 . OpenUrl CrossRef PubMed 31. ↵ Moshe Sniedovich . Dynamic Programming: Foundations and Principles , Second Edition. CRC Press , 2 edition. ISBN 978-0-429-11620-9 . doi: 10.1201/EBK0824740993 . OpenUrl CrossRef 32. ↵ João Victor da Silva Guerra , Helder Veras Ribeiro-Filho , Gabriel Ernesto Jara , Lean-dro Oliveira Bortot , José Geraldo de Carvalho Pereira , and Paulo Sérgio Lopes-de Oliveira . pyKVFinder: an efficient and integrable python package for biomolecular cavity detec-tion and characterization in data science . 22 ( 1 ): 607 . ISSN 1471-2105 . doi: 10.1186/s12859-021-04519-4 . OpenUrl CrossRef 33. ↵ Min Fey Chek , Sun-Yong Kim , Tomoyuki Mori , Hasni Arsad , Mohammed Razip Samian , Kumar Sudesh , and Toshio Hakoshima . Structure of polyhydroxyalkanoate (PHA) synthase PhaC from Chromobacterium sp. USM2, producing biodegradable plastics . Scientific Re-ports , 7 ( 1 ): 5312 , July 2017 . ISSN 2045-2322 . doi: 10.1038/s41598-017-05509-4 . Publisher: Nature Publishing Group . OpenUrl CrossRef PubMed 34. ↵ Seongjoon Joo , In Jin Cho , Hogyun Seo , Hyeoncheol Francis Son , Hye-Young Sagong , Tae Joo Shin , So Young Choi , Sang Yup Lee , and Kyung-Jin Kim . Structural insight into molecular mechanism of poly(ethylene terephthalate) degradation . Nature Communica-tions , 9 ( 1 ): 382 , January 2018 . ISSN 2041-1723 . doi: 10.1038/s41467-018-02881-1 . Pub-lisher: Nature Publishing Group . OpenUrl CrossRef 35. ↵ Brandon C. Knott , Erika Erickson , Mark D. Allen , Japheth E. Gado , Rosie Graham , Fiona L. Kearns , Isabel Pardo , Ece Topuzlu , Jared J. Anderson , Harry P. Austin , Graham Do-minick , Christopher W. Johnson , Nicholas A. Rorrer , Caralyn J. Szostkiewicz , Valérie Copié , Christina M. Payne , H. Lee Woodcock , Bryon S. Donohoe , Gregg T. Beckham , and John E. McGeehan . Characterization and engineering of a two-enzyme system for plastics de-polymerization . Proceedings of the National Academy of Sciences , 117 ( 41 ): 25476 – 25485 , 2020 . doi: 10.1073/pnas.2006753117 . OpenUrl Abstract / FREE Full Text 36. ↵ Charles Eigenbrot . Structure-function of EGFR kinase domain and its inhibitors , pages 30 – 44 . Humana Press , Totowa, NJ , 2008 . ISBN 978-1-59745-356-1 . doi: 10.1007/978-1-59745-356-1_3 . OpenUrl CrossRef 37. ↵ Søren G. F. Rasmussen , Hee-Jung Choi , Daniel M. Rosenbaum , Tong Sun Kobilka , Foon Sun Thian , Patricia C. Edwards , Manfred Burghammer , Venkata R. P. Ratnala , Ruslan Sanishvili , Robert F. Fischetti , Gebhard F. X. Schertler , William I. Weis , and Brian K. Ko-bilka . Crystal structure of the human β2 adrenergic g-protein-coupled receptor . 450 ( 7168 ): 383 – 387 . ISSN 1476-4687 . doi: 10.1038/nature06325 . Publisher: Nature Publishing Group . OpenUrl CrossRef PubMed Web of Science 38. Castillo et al. | ActSeek: Fast and accurate search algorithm of active sites bioRχiv | 7 View the discussion thread. Back to top Previous Next Posted February 15, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following ActSeek: Fast and accurate search algorithm of active sites in Alphafold database Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share ActSeek: Fast and accurate search algorithm of active sites in Alphafold database Sandra Castillo , O. H. Samuli Ollila bioRxiv 2025.02.11.637678; doi: https://doi.org/10.1101/2025.02.11.637678 Share This Article: Copy Citation Tools ActSeek: Fast and accurate search algorithm of active sites in Alphafold database Sandra Castillo , O. H. Samuli Ollila bioRxiv 2025.02.11.637678; doi: https://doi.org/10.1101/2025.02.11.637678 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Bioinformatics Subject Areas All Articles Animal Behavior and Cognition (7635) Biochemistry (17697) Bioengineering (13894) Bioinformatics (41951) Biophysics (21456) Cancer Biology (18594) Cell Biology (25515) Clinical Trials (138) Developmental Biology (13380) Ecology (19903) Epidemiology (2067) Evolutionary Biology (24322) Genetics (15612) Genomics (22510) Immunology (17737) Microbiology (40400) Molecular Biology (17183) Neuroscience (88619) Paleontology (667) Pathology (2833) Pharmacology and Toxicology (4825) Physiology (7644) Plant Biology (15158) Scientific Communication and Education (2046) Synthetic Biology (4296) Systems Biology (9825) Zoology (2271)