

research papers

FLEXR-MSA: electron-density map comparisons of sequence-diverse structures
aDepartment of Chemical Biology and Therapeutics, MS 1000, St Jude Children's Research Hospital, 262 Danny Thomas Place, Memphis, TN 38105, USA
*Correspondence e-mail: marcus.fischer@stjude.org
Proteins with near-identical sequences often share similar static structures. Yet, comparing crystal structures is limited or even biased by what has been included or omitted in the deposited model. Information about unique dynamics is often hidden in electron-density maps. Currently, automatic map comparisons are limited to sequence-identical structures. To overcome this limitation, we developed FLEXR-MSA, which enables unbiased electron-density map comparisons of sequence-diverse structures by coupling multiple sequence alignment (MSA) with electron-density sampling. FLEXR-MSA generates visualizations that pinpoint low-occupancy features on the residue level and chart them across the protein surface to reveal global changes. To exemplify the utility of this tool, we probed electron densities for protein-wide alternative conformations of HSP90 across four human isoforms and other homologs. Our analysis demonstrates that FLEXR-MSA can reveal hidden differences among HSP90 variants bound to clinically important ligands. Integrating this new functionality into the FLEXR suite of tools links the comparison of conformational landscapes hidden in electron-density maps to the building of multi-conformer models that reveal structural/functional differences that might be of interest when designing selective ligands.
Keywords: electron-density sampling; protein conformational landscape; HSP90; isoforms; dynamics; ligand discovery.
1. Introduction
Proteins are peripatetic (Matthews, 2010), so that at each point in time they exist as a collection of major and minor states. In response to perturbations such as ligand binding (Merski et al., 2015
; Wankowicz et al., 2022
; Stachowski & Fischer, 2022
) or temperature (Fischer, 2021
; Fischer et al., 2015
; Keedy, 2019
; Stachowski et al., 2022
) the relative populations of these states are reshaped (Frauenfelder et al., 1991
; Henzler-Wildman & Kern, 2007
; Yabukarski, Doukov, Mokhtari et al., 2022
). This flexibility is essential to many functions, including enzyme catalysis and membrane transport. Detecting areas of flexibility can reveal new opportunities for developing biological or technical advances (Bradford et al., 2021
; Fischer et al., 2014
; Aplin et al., 2022
; Yabukarski, Doukov, Pinney et al., 2022
).
Despite the recent `resolution revolution' in cryo-electron microscopy (Kühlbrandt, 2014), X-ray crystallography is still the most popular tool for determining near-atomic resolution protein structures. Crystallographic electron-density maps solved to sufficient resolution contain information about dynamics such as weakly populated and high-energy minor states (Fraser et al., 2011
; Lang et al., 2010
; Pearce et al., 2017
; Pearce & Gros, 2021
). Estimates from retrospective analyses of deposited X-ray data suggest that up to a third of protein side chains show evidence of minor states in electron-density maps but are not accounted for in the corresponding models (Fraser et al., 2011
; Bradford et al., 2021
; Shapovalov & Dunbrack, 2007
; Lang et al., 2010
). The incompleteness of published models is partially due to the fact that the signal for flexible features such as alternative conformations of side chains is often weak. This makes it challenging to accurately discern genuine signal from experimental noise. Additionally, incoporating conformational dynamics into models necessitates manual intervention, which is cumbersome and inaccessible for non-crystallographers. Emerging automated multi-state modeling tools such as qFit, Phenix-MD and FLEXR try to bridge this gap (Riley et al., 2021
; Burnley et al., 2012
; Stachowski & Fischer, 2023
, 2024
). Electron-density measurements without explicit model building using tools such as Ringer (Lang et al., 2010
) circumvent these pitfalls and allow the visualization of side-chain dynamics without modeling bias. Current wisdom supports that alternate side chains can be confidently interpreted in weak electron density (≳0.3σ; Lang et al., 2010
), which enables older maps deposited at a time of more cautious modeling guidelines (previously >1σ) to be searched. While ensemble methods build comprehensive models, comparing structural differences between proteins with non-identical sequences remains challenging.
One of the cornerstone approaches for probing the protein conformational landscape is through mutagenesis (Winter et al., 1982), where structural and functional consequences are monitored when substituting amino acids with different properties (Fowler & Fields, 2014
). Nature took advantage of this to develop highly specialized proteins from related ones, for example through sequence divergence (Chothia & Lesk, 1986
) or alternative splicing (Baralle & Giudice, 2017
). However, nature's ingenuity creates a large hurdle for drug discovery. Poor selectivity of sequence-related but functionally distinct proteins often leads to serious off-target effects for clinical targets such as human histone deacetylase (Ma et al., 2016
), carbonic anhydrase (Alterio et al., 2012
), kinases (Ferguson & Gray, 2018
) and bromodomains (Liu et al., 2017
). Generally, aspects of protein flexibility can be used to improve ligand affinity and selectivity (Teague, 2003
).
Another well known example is the heat-shock protein 90 (HSP90) family of molecular chaperones. HSP90 proteins drive all ten hallmarks of cancer (Hanahan & Weinberg, 2011; Garg et al., 2016
) but no inhibitor has been clinically approved outside Japan (Yuno et al., 2018
). Humans possess four HSP90 isoforms (Hsp90α, Hsp90β, Grp94 and Trap1) that share greater than 90% sequence identity in the N-terminal domain (NTD) binding site alone, where Hsp90α and Hsp90β differ by only two residues (Stachowski et al., 2023
; Supplementary Fig. S1). An isoform-selective inhibitor is a promising avenue to avoid inducing the cellular heat-shock response and eventual tumor resistance (Mishra et al., 2021
; Huck et al., 2019
; Ernst et al., 2014
; Gewirth, 2016
). Likewise, Hsp90α is targeted in antifungal drug development, but the close similarity between the human and fungal homologs causes severe host toxicities (Cowen et al., 2009
; Supplementary Fig. S2). Candida albicans is the most common fungal pathogen affecting humans. While C. albicans Hsp90α shares 72% sequence identity with the human homolog NTD, the binding site remains largely conserved with only two residues changing: S52A and V186L (according to the human sequence numbering; Supplementary Figs. S1 and S2). Despite their similar sequences, there are major structural differences in ligand binding between the C. albicans and human homologs that might open routes for developing targeted antifungal therapies (Whitesell et al., 2019
). These differences primarily include rearrangements in the ATP lid-loop region, which is known to be highly dynamic and ligand-responsive in the human form (Amaral et al., 2017
; Stachowski & Fischer, 2022
) but possibly more so in C. albicans (Whitesell et al., 2019
). With HSP90 proteins being remarkly flexible (Stachowski & Fischer, 2022
) and the human isoforms exhibiting subtle but meaningful structural differences, this opens new routes for (Khandelwal et al., 2018
; Huck et al., 2019
; Stachowski et al., 2023
).
Here, we combine electron-density map sampling with multiple sequence alignment (MSA) into FLEXR-MSA as a tool for comparing the electron densities of structures with mutations, dissimilar sequences and misnumbered residues. For HSP90, this tool enabled us to directly probe electron-density maps for protein-wide alternative side-chain conformations across three homologs. More generally, our analysis demonstrates that FLEXR-MSA can offer new insights into structural differences among sequence-dissimilar proteins that are often missed in static models. The tool is open source and is available within FLEXR on GitHub at https://github.com/TheFischerLab/FLEXR.
2. Materials and methods
Coordinates and structure factors (Supplementary Tables S1 and S2) were taken from the Protein Data Bank (PDB; Berman et al., 2000). For Hsp90α, we compared structures according to Whitesell and coworkers except in the case of the apo human structure (Supplementary Table S1; Whitesell et al., 2019
), where the authors used PDB entry 1yer, which was deposited without structure factors. We used PDB entry 1uyl, which is also apo, solved at a comparable resolution (1.7 Å for PDB entry 1yer and 1.4 Å for PDB entry 1uyl) and has the same lid conformation (`in'). Maps were examined with Ringer as described previously (Lang et al., 2010
). PyMOL (Schrödinger, New York, USA) was used to generate images and to detect conformational changes in the ATP lid. All-atom r.m.s.d. values and structural superpositions were also performed in PyMOL using align mobile.pdb, target.pdb, cycles=0. These structure-based alignments are not considered in FLEXR-MSA. Chains in structures with multiple copies were treated as separate models, except in the case of PDB entry 3opd where, due to the lower resolution (2.6 Å), only the A chain was considered. Binding-site volumes and hydrophiblic–hydrophobic balance were calculated with SiteMap (Halgren, 2009
) in Maestro (Schrödinger, New York, USA).
FLEXR-MSA was written in Python 3.9 and packaged within the FLEXR suite of tools. Full functionality of FLEXR, including the GUI (Stachowski & Fischer, 2024), requires Coot 1.1.10 (Emsley, 2023
), which we recommend installing through CCP4 version 9 (Agirre et al., 2023
). FLEXR is available as an open-source program in a GitHub repository (https://github.com/TheFischerLab/FLEXR) and requires the Biopython, Matplotlib, Numpy, Pandas and SciPy Python packages. Ringer is available in the mmtbx library (https://cctbx.github.io/mmtbx/mmtbx.html) or in Phenix (Liebschner et al., 2019
). MUSCLE version 5.2 (Edgar, 2004
) is also available through Homebrew (https://github.com/brewsci/homebrew-bio/blob/develop/Formula/muscle.rb) or can be installed separately (https://www.drive5.com/muscle). Ringer peak detection and peak subtraction were performed as described previously (Stachowski et al., 2022
). Pearson (CC) calculations were performed with the SciPy Python package (Virtanen et al., 2020
). Surface visualizations require PyMOL. A detailed protocol for running FLEXR-MSA is given in the Supplementary Methods.
3. Results
3.1. Program description
The FLEXR-MSA workflow is illustrated in Fig. 1. After the user runs Ringer, FLEXR-MSA starts from the standard Ringer CSV output files that contain σ measurements taken around each dihedral angle (χ) for each except Gly and Ala, in a PDB structure (Lang et al., 2010
). The amino-acid sequence is extracted from the Ringer output and organized into FASTA format. A multiple sequence alignment (MSA) is performed with MUSCLE (Edgar, 2004
). Residues are renumbered according to their location in the MSA, and their relation to the numbering in the input Ringer CSVs are saved in a look-up table. To produce classical `Ringer plots' (σ values as a function of side-chain rotation angle) for each residue, σ values are extracted at each position in the alignment for each sequence. These image files are saved to the working directory; the plot title and file name correspond to the MSA position. This process is repeated for each χ angle. To facilitate quick cross-comparison the original PDB residue number and chain ID is captured in the figure legend (see Fig. 1
). The alignment files are also saved and can be manually adjusted and reloaded. Colors can be defined by the user and otherwise are automatically assigned (see Supplementary Methods). Median Pearson CC values are calculated and saved in the B-factor column of a given PDB file to be visualized in PyMOL. Starting from the Ringer output, the whole process takes less than a minute for these HSP90 comparisons.
![]() | Figure 1 FLEXR-MSA workflow for comparison of alternative side-chain conformations in electron-density maps across related, sequence-diverse proteins. (1) FLEXR-MSA reads in the CSV output file with the σ values from Ringer (Lang et al., 2010 ![]() ![]() |
3.2. Detecting alternative conformations across isoforms
To illustrate the utility of FLEXR-MSA, we chose the structurally dynamic HSP90 family of molecular chaperones. The high sequence identity and structural similarity among its four human isoforms has made it difficult to discover isoform-selective compounds. We applied FLEXR-MSA to structures of each isoform bound to the same fragment, N,N-dimethyl-7H-purin-6-amine (6DMP; PDB ID 42C; Stachowski et al., 2023). 6DMP contains the core purine scaffold that is present in the native substrate ATP and is a common starting point in ligand discovery.
To find changes that may impact ligand binding, we focused on binding-site residues. All isoforms contain a conserved Asp that is often exploited to hydrogen-bond to ligands (Chiosis et al., 2001). This Asp is surrounded by a conserved water network that varies in position due to the loss of a hydrogen bond from a nearby mutation from Ser in Hsp90α to Ala in Hsp90β, Grp94 and Trap1. This distinguishing feature was previously exploited to design ligands that displace or retain certain waters and improve α/β selectivity (Khandelwal et al., 2018
; Mishra et al., 2021
; Huck et al., 2019
). Here, all isoforms share the same predominate conformation of the Asp [Fig. 2
(a)]. However, FLEXR-MSA reveals that two of the four chains in Hsp90β contain an additional Asp rotamer that is not present in the other isoforms (A at ∼340° and D at ∼190°) [Fig. 2
(b)]. It is conceivable that the additional conformations may be facilitated by the greater flexibility of the water network in Hsp90β over Hsp90α due to the Ser-to-Ala mutation that differentiates the two cytoplasmic isoforms.
![]() | Figure 2 FLEXR-MSA reveals previously unnoticed isoform-specific conformations across all four 6DMP-bound human HSP90s. (a) Hsp90α (green), Hsp90β (blue), Grp94 (yellow) and Trap1 (red) bound to 6DMP (inset, PDB ligand ID 42C). (b) Ringer plot produced by FLEXR-MSA showing additional conformations (arrows) of a conserved binding-site Asp in Hsp90β chains A (∼340°) and D (∼190°). The solid line corresponds to 0σ. The gray dashed line corresponds to 0.3σ, the Ringer cutoff for modeling. The black dashed line corresponds to 1σ, the conventional modeling threshold. |
3.3. Detecting specific conformations between human and C. albicans Hsp90α
To better understand homolog-specific flexibility in ligand binding, we used FLEXR-MSA to reanalyze four pairs of human and C. albicans Hsp90α structures: one apo and three bound to matching ligands first reported by Whitesell et al. (2019) (Supplementary Table S2).
First, we inspected the binding site in apo C. albicans and human structures. FLEXR-MSA revealed that the apo electron-density map for human Hsp90α shows an alternative conformation of a conserved methionine that is not present in the C. albicans map [Fig. 3(a)]. The origin of this change in the population of Met98/87 (human/C. albicans numbering) conformations might be a consequence of the different position of the lid, which is in the `in' conformation in the human protein and the `out' conformation in that from C. albicans (r.m.s.d. of 1.5 Å; Corbett & Berger, 2010
). The alternative conformation repositions the terminal sulfur–carbon group of Met98/87. As a consequence of these conformational differences the binding site shifts its hydrophilic–hydrophobic balance towards more hydrophobic (0.71 in human and 0.47 in C. albicans). This change provides different surfaces to target, although the binding-site volume change may appear to be negligible (280 Å3 in the human protein versus 277 Å3 in that from C. albicans).
![]() | Figure 3 Homolog-specific HSP90 conformations. (a) Comparison of human (blue) and C. albicans (red) apo Hsp90α shows an additional conformation of a binding-site Met in the Ringer plot for the human protein. (b) Comparison of human and C. albicans Hsp90α bound to AUY-922 shows an additional conformation of a Ser in the C. albicans protein in the inset Ringer plot. (c) Radicicol (RDC) bound to C. albicans Hsp90α has an additional conformation of a conserved Asp in the C. albicans form, while Lys58/47 is more variable in the human form (d, e). (e) Comparison of SNX-2112 bound to human, C. albicans and T. brucei (purple) Hsp90α. The A and B chains are shown for human and C. albicans and chain A is shown for T. brucei. The inset Ringer plot shows that the Lys is conformationally variable between homologs bound to the conformation-responsive SNX-2112 ligand. Dotted lines represent hydrogen bonds detected with PyMOL. 2Fo − Fc maps are shown in blue and contoured at 1σ. Fo − Fc maps are shown in red and green and contoured at ±2σ. |
Secondly, we were interested in understanding the impact of sequence differences on binding AUY-922 (luminespib), which is an experimental drug candidate that reached Phase II (Felip et al., 2018) in clinical trials for several cancer types. Binding of AUY-922 leads to different protein and ligand conformations between human and C. albicans Hsp90α. The lid in the human–AUY-922 complex is in the `in' state, mirroring the apo conformation, while the lid in C. albicans is in the `helical' state (lid r.m.s.d. of 3.6 Å; Whitesell et al., 2019
; Supplementary Fig. S3). This change in lid state repositions the terminal morpholine substituent and leads to different polar interactions, with an overall ligand r.m.s.d. of 1.4 Å. Differences in ligand position cascade throughout the binding site and reposition water molecules and proximal unengaged residues such as Ser50/39, which has an additional conformation in C. albicans [Fig. 3
(b)]. In the newly identified conformation, the Ser hydroxyl points away from the binding site. This indicates that it might be a less accessible interaction partner for ligand binding than suggested by the original single-conformer model.
Thirdly, radicicol (RDC) is a potent macrocyclic inhibitor of HSP90-dependent tumor growth (Roe et al., 1999). The overall fold between both the human and C. albicans Hsp90α structure (r.m.s.d.s of 1.4 Å for chain A and 1.2 Å for chain B) and RDC pose (r.m.s.d. of 0.14 Å for both chains) are similar. In both homologs, RDC forms hydrogen bonds with the conserved residue Asp93/82 [Fig. 3
(c)]. However, our analysis revealed a second high-energy conformation of Asp93/82 in the C. albicans structure. Notably, in the human structure the dynamic Lys58/47 engages with RDC, while no direct interactions are formed in the C. albicans structure. Using FLEXR-MSA we detected a second weak conformation of this Lys in the electron-density maps of the human form that points away from RDC [Fig. 3
(d)].
3.4. Detecting conformational differences in HSP90 across three homologs
Next, we expanded our comparison from human and C. albicans to a third homolog by considering HSP90 from the parasitic protist Trypanosoma brucei bound to SNX-2112. There is a considerable rearrangement of `helical' residues Val93–Ser102 in the T. brucei and C. albicans forms bound to SNX-2112 that is absent in the human form (Whitesell et al., 2019; Supplementary Fig. S3). These differences within the same lid state were proposed to contribute to the variability in affinities between homologs (Whitesell et al., 2019
). Our analysis detected additional homolog-specific states away from the lid site that might allosterically modulate affinity. Specifically, we identified that the ligand-responsive Lys58/47 in RDC structures [Figs. 3
(d)] also shifts conformation between homologs on binding SNX-2112 [Fig. 3
(e)]. In both chains of the human structure, this Lys is in a consistent position and hydrogen-bonds to SNX-2112. The C. albicans structure contains two protein copies but only one chain is occupied by the ligand. The Lys in the bound chain is in a different conformation than in the human protein but remains hydrogen-bonded to the ligand, which shifts by an r.m.s.d. of 2.4 Å between the human and C. albicans structures. In contrast to the bound chain, this Lys points away from the binding site in the apo C. albicans chain. Interestingly, in the T. brucei structure the Lys is in a mixture of both the bound and unbound conformations from the C. albicans structure and results in a different ligand pose. The ring featuring the oxygen closest to the Lys is most affected (r.m.s.d.s of 2.8 Å to the human structure and 2.5 Å to that from C. albicans).
3.5. Mapping global, homolog-specific conformational differences
To quantify how many side-chain conformations change between human and C. albicans Hsp90α–RDC we subtracted the number of peaks in aligned Ringer plots [Figs. 4(a) and 4
(b)]. Mapping these peak-count differences onto the protein surface allowed us to pinpoint local conformational differences in human and C. albicans RDC-bound structures. This revealed that several conserved residues in the RDC binding site in C. albicans HSP90 have additional conformations that are not present in the human structure. In contrast, flexibility in human HSP90 is greater for residues along the lid. While these residues do not directly interact with RDC, they provide ligand access to the pocket and often reposition upon binding ligands of different chemotypes (Stachowski & Fischer, 2022
) [Figs. 4
(a) and 4
(b)]. Pearson CC values between aligned Ringer plots support this as well: while the canonical nucleotide-binding site is positioned similarly between homologs, the ATP lid is variable [Figs. 4
(c) and 4
(d), Supplementary Fig. S3]. Mapping these values across the entire protein surface shows additional regions with varying dynamics both near and far from the orthosteric binding site.
![]() | Figure 4 Mapping homolog-specific regional conformational variability in HSP90 bound to RDC. Difference in χ1 Ringer peaks between human and C. albicans Hsp90α–RDC mapped onto the front (a) and back (b) of human Hsp90α–RDC. The red surface coloring corresponds to more alternative conformers as peaks in electron-density maps in the C. albicans protein and the blue surface corresponds to more peaks in human Hsp90α bound to RDC (shown as yellow sticks). Ringer plots are shown for binding-site residues with alternative conformations between homologs. This is also supported by mapping Pearson CC values for χ1 (c) and χ2 (d) onto the protein surface. |
4. Discussion
At sufficient resolution, electron-density maps often contain details that describe protein dynamics that are missing in the deposited structural models. To enable an unbiased comparison of hidden, low-occupancy features in electron-density maps across diverse proteins, we combined electron-density sampling with MSA. We used this tool, FLEXR-MSA, to compare alternative side-chain conformations in electron-density maps of HSP90 across four human isoforms, a fungal homolog and a protist homolog.
Three main implications for ligand binding emerge from this work. Firstly, FLEXR-MSA revealed changes in the binding-site conformations of four human HSP90 isoforms, beyond obvious sequence dissimilarities, that were hiding in the electron-density maps. Secondly, we found homolog-specific conformational variability of charged residues in comparisons of three Hsp90α homologs bound to varying ligands. This was most profound for a conserved Lys that changed conformation with ligand pose between homologs and ligand-bound states. Thirdly, a protein-wide comparison of human and C. albicans Hsp90α bound to RDC showed different orthosteric and potential allosteric regions of heightened variability between homologs.
It is worth keeping in mind that even near-identical proteins differ in their conformational landscape. For instance, binding-site conformations are often connected to water networks, so that repopulating side chains will shift water networks and vice versa (Darby et al., 2019). Likewise, differences in the amino-acid sequence alter water-network connectivity even if waters within the network are conserved. Taking advantage of this phenomenon to displace specific waters in HSP90 has provided an interesting route to selectively target the α and β isoforms (Khandelwal et al., 2018
; Mishra et al., 2021
; Huck et al., 2019
). Here, we observed additional weak conformations of a conserved Asp (Asp88) in two of the four Hsp90β chains. This change in populations might be an underappreciated consequence of the change from an adjacent Ala to Ser between α and β. In recent work we showed that waters in HSP90 isoforms bound to the same ligand, 6DMP, exhibited distinct behaviors regarding r.m.s.d. and normalized B factors (Stachowski et al., 2023
). This included waters that bridge interactions between 6DMP and this conserved Asp. In another example, changed lid states in Hsp90α bound to AUY-922 between the human and C. albicans proteins cascade through the binding site and reposition ligands, water networks and side chains. Connecting weakly populated states observed here with changes in water behavior and sequence differences might provide new insights to selectively target HSP90 homologs.
When trying to understand homolog-specific differences, the focus is generally on sequence differences. Here, we have illustrated the ability of FLEXR-MSA to detect homolog-selective repositioning of two conserved charged residues Lys (58 in α) and Asp (93 in α) in the nucleotide binding site. For instance, in the case of C. albicans Hsp90α bound to SNX-2112 the Lys exists in two distinct conformations between ligand-bound and unbound states. Consequently, the ligand moiety interacting with the Lys also varied with the Lys conformation while the rest of the pose was conserved across homologs. This same Lys in the human homologue was observed to be temperature-sensitive (Stachowski et al., 2022) and a selective handle for covalent inhibitor design (Cuesta et al., 2020
). Whitesell and coworkers reported that SNX-2112 exhibited a threefold higher affinity for C. albicans Hsp90α over human (Whitesell et al., 2019
) and others reported a higher affinity for the T. brucei protein over human (Pizarro et al., 2013
). AUY-922 also exhibited higher affinity for the human form compared with that from C. albicans (Whitesell et al., 2019
). Hidden changes in homolog-specific flexibilities might explain some of these differences in affinities.
While the FLEXR-MSA approach facilitates the inspection of electron-density maps for alternative side-chain conformations in sequence-diverse proteins, the responsibility for sensible data input and cautious analysis is still with the user (Pozharski et al., 2013). Users need to carefully consider other influences on structure such as resolution and crystallization and experimental conditions. For instance, to test the consistency of these observations we treated chains as separate lines of evidence when possible. Differences in conformations between chains could result from varying microenvironments within the However, with careful consideration that specific rotamers are not distorted involuntarily, the presence of conformational heterogeneity alone can be enlightening. Also, differences in resolution will create different thresholds for detecting high-energy, weakly populated states. For instance, the additional Asp conformations in β (1.8 Å) were not present in Grp94 (2.3 Å) or Trap1 (2.3 Å) although all three share the adjacent Ala substitution in lieu of Ser in α. We cannot rule out that rare Asp conformations are absent due to the reduced signal-to-noise ratio at the lower resolution of the Grp94 and Trap1 structures. The best way to validate these observations is to add the conformations into the model, for example with FLEXR, re-refine the structure against the diffraction data and monitor occupancies and clashes (Stachowski & Fischer, 2023
, 2024
).
Also, the robustness of the FLEXR-MSA approach is directly dependent on the success of the sequence alignment, which in turn is linked to sequence similarity and the completeness of sequences as they are extracted from the Ringer output, which excludes residues without χ angles, such as Ala and Gly, or unbuilt portions of the model. Inherently sequence alignments can be quite poor at the beginning and end of chains and adjacent to unbuilt loops. However, these regions also typically correspond to areas of weak electron density and this lack of signal reduces confidence in any observation in that region of the protein so that analysis may not be useful. To overcome this potential limitation, FLEXR-MSA saves alignment and re-indexing files that can be referenced and modified. If the alignment approach appears to be limiting, FLEXR-MSA allows the users to manually change the alignment and MUSCLE contains several options that can additionally be adjusted to improve the alignment (Edgar, 2004).
FLEXR-MSA was designed for comparing electron densities of structures with mutations, dissimilar sequences and misnumbered residues. Combing electron-density sampling with MSA bypasses many tedious steps and allows users to quickly visualize and analyze electron-density features of structures with non-identical sequences. FLEXR-MSA is fast, portable and relies only on common dependencies. It is available within the FLEXR suite, and, as such, is freely available to the community.
Supporting information
Supplementary tables and figures, and supplementary methods. DOI: https://doi.org/10.1107/S2052252525001332/jt5079sup1.pdf
Acknowledgements
We thank the High-Performance Computing Facility for ongoing support. We also thank Fatimah Oyedeji for providing feedback on the software and tutorial. Author contributions were as follows. TRS and MF designed the research; TRS performed the research and contributed new analytic tools; TRS and MF analyzed the data and wrote the paper.
Funding information
This work was supported by the American Lebanese Syrian Associated Charities (ALSAC) and National Institute of General Medical Sciences grant R35GM142772 (to MF) and an Academic Programs Special Fellowship (TRS).
References
Agirre, J., Atanasova, M., Bagdonas, H., Ballard, C. B., Baslé, A., Beilsten-Edmands, J., Borges, R. J., Brown, D. G., Burgos-Mármol, J. J., Berrisford, J. M., Bond, P. S., Caballero, I., Catapano, L., Chojnowski, G., Cook, A. G., Cowtan, K. D., Croll, T. I., Debreczeni, J. É., Devenish, N. E., Dodson, E. J., Drevon, T. R., Emsley, P., Evans, G., Evans, P. R., Fando, M., Foadi, J., Fuentes-Montero, L., Garman, E. F., Gerstel, M., Gildea, R. J., Hatti, K., Hekkelman, M. L., Heuser, P., Hoh, S. W., Hough, M. A., Jenkins, H. T., Jiménez, E., Joosten, R. P., Keegan, R. M., Keep, N., Krissinel, E. B., Kolenko, P., Kovalevskiy, O., Lamzin, V. S., Lawson, D. M., Lebedev, A. A., Leslie, A. G. W., Lohkamp, B., Long, F., Malý, M., McCoy, A. J., McNicholas, S. J., Medina, A., Millán, C., Murray, J. W., Murshudov, G. N., Nicholls, R. A., Noble, M. E. M., Oeffner, R., Pannu, N. S., Parkhurst, J. M., Pearce, N., Pereira, J., Perrakis, A., Powell, H. R., Read, R. J., Rigden, D. J., Rochira, W., Sammito, M., Sánchez Rodríguez, F., Sheldrick, G. M., Shelley, K. L., Simkovic, F., Simpkin, A. J., Skubak, P., Sobolev, E., Steiner, R. A., Stevenson, K., Tews, I., Thomas, J. M. H., Thorn, A., Valls, J. T., Uski, V., Usón, I., Vagin, A., Velankar, S., Vollmar, M., Walden, H., Waterman, D., Wilson, K. S., Winn, M. D., Winter, G., Wojdyr, M. & Yamashita, K. (2023). Acta Cryst. D79, 449–461. Web of Science CrossRef IUCr Journals Google Scholar
Alterio, V., Di Fiore, A., D'Ambrosio, K., Supuran, C. T. & De Simone, G. (2012). Chem. Rev. 112, 4421–4468. Web of Science CrossRef CAS PubMed Google Scholar
Amaral, M., Kokh, D. B., Bomke, J., Wegener, A., Buchstaller, H. P., Eggenweiler, H. M., Matias, P., Sirrenberg, C., Wade, R. C. & Frech, M. (2017). Nat. Commun. 8, 2276. Google Scholar
Aplin, C., Milano, S. K., Zielinski, K. A., Pollack, L. & Cerione, R. A. (2022). J. Phys. Chem. B, 126, 6599–6607. Google Scholar
Baralle, F. E. & Giudice, J. (2017). Nat. Rev. Mol. Cell Biol. 18, 437–451. Google Scholar
Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H., Shindyalov, I. N. & Bourne, P. E. (2000). Nucleic Acids Res. 28, 235–242. Web of Science CrossRef PubMed CAS Google Scholar
Bradford, S. Y. C., El Khoury, L., Ge, Y., Osato, M., Mobley, D. L. & Fischer, M. (2021). Chem. Sci. 12, 11275–11293. Web of Science CrossRef CAS PubMed Google Scholar
Burnley, B. T., Afonine, P. V., Adams, P. D. & Gros, P. (2012). eLife, 1, e00311. Web of Science CrossRef PubMed Google Scholar
Chiosis, G., Timaul, M. N., Lucas, B., Munster, P. N., Zheng, F. F., Sepp-Lorenzino, L. & Rosen, N. (2001). Chem. Biol. 8, 289–299. Google Scholar
Chothia, C. & Lesk, A. M. (1986). EMBO J. 5, 823–826. CrossRef CAS PubMed Web of Science Google Scholar
Corbett, K. D. & Berger, J. M. (2010). Proteins, 78, 2738–2744. Google Scholar
Cowen, L. E., Singh, S. D., Köhler, J. R., Collins, C., Zaas, A. K., Schell, W. A., Aziz, H., Mylonakis, E., Perfect, J. R., Whitesell, L. & Lindquist, S. (2009). Proc. Natl Acad. Sci. USA, 106, 2818–2823. Google Scholar
Cuesta, A., Wan, X., Burlingame, A. L. & Taunton, J. (2020). J. Am. Chem. Soc. 142, 3392–3400. Google Scholar
Darby, J. F., Hopkins, A. P., Shimizu, S., Roberts, S. M., Brannigan, J. A., Turkenburg, J. P., Thomas, G. H., Hubbard, R. E. & Fischer, M. (2019). J. Am. Chem. Soc. 141, 15818–15826. Web of Science CrossRef CAS PubMed Google Scholar
Edgar, R. C. (2004). Nucleic Acids Res. 32, 1792–1797. Web of Science CrossRef PubMed CAS Google Scholar
Emsley, P. (2023). Coot. https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/. Google Scholar
Ernst, J. T., Neubert, T., Liu, M., Sperry, S., Zuccola, H., Turnbull, A., Fleck, B., Kargo, W., Woody, L., Chiang, P., Tran, D., Chen, W., Snyder, P., Alcacio, T., Nezami, A., Reynolds, J., Alvi, K., Goulet, L. & Stamos, D. (2014). J. Med. Chem. 57, 3382–3400. Google Scholar
Felip, E., Barlesi, F., Besse, B., Chu, Q., Gandhi, L., Kim, S. W., Carcereny, E., Sequist, L. V., Brunsvig, P., Chouaid, C., Smit, E. F., Groen, H. J. M., Kim, D. W., Park, K., Avsar, E., Szpakowski, S., Akimov, M. & Garon, E. B. (2018). J. Thorac. Oncol. 13, 576–584. Google Scholar
Ferguson, F. M. & Gray, N. S. (2018). Nat. Rev. Drug Discov. 17, 353–377. Google Scholar
Fischer, M. (2021). Q. Rev. Biophys. 54, e1. Web of Science CrossRef PubMed Google Scholar
Fischer, M., Coleman, R. G., Fraser, J. S. & Shoichet, B. K. (2014). Nat. Chem. 6, 575–583. Web of Science CrossRef CAS PubMed Google Scholar
Fischer, M., Shoichet, B. K. & Fraser, J. S. (2015). ChemBioChem, 16, 1560–1564. Web of Science CrossRef CAS PubMed Google Scholar
Fowler, D. M. & Fields, S. (2014). Nat. Methods, 11, 801–807. Google Scholar
Fraser, J. S., van den Bedem, H., Samelson, A. J., Lang, P. T., Holton, J. M., Echols, N. & Alber, T. (2011). Proc. Natl Acad. Sci. USA, 108, 16247–16252. Web of Science CrossRef CAS PubMed Google Scholar
Frauenfelder, H., Sligar, S. G. & Wolynes, P. G. (1991). Science, 254, 1598–1603. CrossRef PubMed CAS Web of Science Google Scholar
Garg, G., Khandelwal, A. & Blagg, B. S. (2016). Adv. Cancer Res. 129, 51–88. Web of Science CrossRef CAS PubMed Google Scholar
Gewirth, D. T. (2016). Curr. Top. Med. Chem. 16, 2779–2791. Google Scholar
Halgren, T. A. (2009). J. Chem. Inf. Model. 49, 377–389. Web of Science CrossRef PubMed CAS Google Scholar
Hanahan, D. & Weinberg, R. A. (2011). Cell, 144, 646–674. Web of Science CrossRef CAS PubMed Google Scholar
Henzler-Wildman, K. & Kern, D. (2007). Nature, 450, 964–972. Web of Science PubMed CAS Google Scholar
Huck, J. D., Que, N. L. S., Sharma, S., Taldone, T., Chiosis, G. & Gewirth, D. T. (2019). Proteins, 87, 869–877. Google Scholar
Keedy, D. A. (2019). Acta Cryst. D75, 123–137. Web of Science CrossRef IUCr Journals Google Scholar
Khandelwal, A., Kent, C. N., Balch, M., Peng, S., Mishra, S. J., Deng, J., Day, V. W., Liu, W., Subramanian, C., Cohen, M., Holzbeierlein, J. M., Matts, R. & Blagg, B. S. J. (2018). Nat. Commun. 9, 425. Web of Science CrossRef PubMed Google Scholar
Kühlbrandt, W. (2014). Science, 343, 1443–1444. Web of Science PubMed Google Scholar
Lang, P. T., Ng, H. L., Fraser, J. S., Corn, J. E., Echols, N., Sales, M., Holton, J. M. & Alber, T. (2010). Protein Sci. 19, 1420–1431. Web of Science CrossRef CAS PubMed Google Scholar
Liebschner, D., Afonine, P. V., Baker, M. L., Bunkóczi, G., Chen, V. B., Croll, T. I., Hintze, B., Hung, L.-W., Jain, S., McCoy, A. J., Moriarty, N. W., Oeffner, R. D., Poon, B. K., Prisant, M. G., Read, R. J., Richardson, J. S., Richardson, D. C., Sammito, M. D., Sobolev, O. V., Stockwell, D. H., Terwilliger, T. C., Urzhumtsev, A. G., Videau, L. L., Williams, C. J. & Adams, P. D. (2019). Acta Cryst. D75, 861–877. Web of Science CrossRef IUCr Journals Google Scholar
Liu, Z., Wang, P., Chen, H., Wold, E. A., Tian, B., Brasier, A. R. & Zhou, J. (2017). J. Med. Chem. 60, 4533–4558. Web of Science CrossRef CAS PubMed Google Scholar
Ma, N., Luo, Y., Wang, Y., Liao, C., Ye, W. C. & Jiang, S. (2016). Curr. Top. Med. Chem. 16, 415–426. Google Scholar
Matthews, B. W. (2010). Protein Sci. 19, 1279–1280. Google Scholar
Merski, M., Fischer, M., Balius, T. E., Eidam, O. & Shoichet, B. K. (2015). Proc. Natl Acad. Sci. USA, 112, 5039–5044. Google Scholar
Mishra, S. J., Khandelwal, A., Banerjee, M., Balch, M., Peng, S., Davis, R. E., Merfeld, T., Munthali, V., Deng, J., Matts, R. L. & Blagg, B. S. J. (2021). Angew. Chem. 133, 10641–10645. Google Scholar
Pearce, N. M. & Gros, P. (2021). Nat. Commun. 12, 5493. Web of Science CrossRef PubMed Google Scholar
Pearce, N. M., Krojer, T., Bradley, A. R., Collins, P., Nowak, R. P., Talon, R., Marsden, B. D., Kelm, S., Shi, J., Deane, C. M. & von Delft, F. (2017). Nat. Commun. 8, 15123. Web of Science CrossRef PubMed Google Scholar
Pizarro, J. C., Hills, T., Senisterra, G., Wernimont, A. K., Mackenzie, C., Norcross, N. R., Ferguson, M. A., Wyatt, P. G., Gilbert, I. H. & Hui, R. (2013). PLoS Negl. Trop. Dis. 7, e2492. Google Scholar
Pozharski, E., Weichenberger, C. X. & Rupp, B. (2013). Acta Cryst. D69, 150–167. Web of Science CrossRef CAS IUCr Journals Google Scholar
Riley, B. T., Wankowicz, S. A., de Oliveira, S. H. P., van Zundert, G. C. P., Hogan, D. W., Fraser, J. S., Keedy, D. A. & van den Bedem, H. (2021). Protein Sci. 30, 270–285. Web of Science CrossRef CAS PubMed Google Scholar
Roe, S. M., Prodromou, C., O'Brien, R., Ladbury, J. E., Piper, P. W. & Pearl, L. H. (1999). J. Med. Chem. 42, 260–266. CrossRef CAS PubMed Google Scholar
Shapovalov, M. V. & Dunbrack, R. L. Jr (2007). Proteins, 66, 279–303. Web of Science CrossRef PubMed CAS Google Scholar
Stachowski, T. R. & Fischer, M. (2022). J. Med. Chem. 65, 13692–13704. Web of Science CrossRef CAS PubMed Google Scholar
Stachowski, T. R. & Fischer, M. (2023). Acta Cryst. D79, 354–367. Web of Science CrossRef IUCr Journals Google Scholar
Stachowski, T. R. & Fischer, M. (2024). J. Appl. Cryst. 57, 580–586. Google Scholar
Stachowski, T. R., Nithianantham, S., Vanarotti, M., Lopez, K. & Fischer, M. (2023). Protein Sci. 32, e4629. Google Scholar
Stachowski, T. R., Vanarotti, M., Seetharaman, J., Lopez, K. & Fischer, M. (2022). Angew. Chem. Int. Ed. 61, e202112919. Web of Science CrossRef Google Scholar
Teague, S. J. (2003). Nat. Rev. Drug Discov. 2, 527–541. Web of Science CrossRef PubMed CAS Google Scholar
Virtanen, P., Gommers, R., Oliphant, T. E., Haberland, M., Reddy, T., Cournapeau, D., Burovski, E., Peterson, P., Weckesser, W., Bright, J., van der Walt, S. J., Brett, M., Wilson, J., Millman, K. J., Mayorov, N., Nelson, A. R. J., Jones, E., Kern, R., Larson, E., Carey, C. J., Polat, I., Feng, Y., Moore, E. W., VanderPlas, J., Laxalde, D., Perktold, J., Cimrman, R., Henriksen, I., Quintero, E. A., Harris, C. R., Archibald, A. M., Ribeiro, A. H., Pedregosa, F., van Mulbregt, P. & SciPy 1.0 Contributors (2020). Nat. Methods, 17, 352. Google Scholar
Wankowicz, S. A., de Oliveira, S. H., Hogan, D. W., van den Bedem, H. & Fraser, J. S. (2022). eLife, 11, e74114. Web of Science CrossRef PubMed Google Scholar
Whitesell, L., Robbins, N., Huang, D. S., McLellan, C. A., Shekhar-Guturja, T., LeBlanc, E. V., Nation, C. S., Hui, R., Hutchinson, A., Collins, C., Chatterjee, S., Trilles, R., Xie, J. L., Krysan, D. J., Lindquist, S., Porco, J. A. Jr, Tatu, U., Brown, L. E., Pizarro, J. & Cowen, L. E. (2019). Nat. Commun. 10, 402. CrossRef PubMed Google Scholar
Winter, G., Fersht, A. R., Wilkinson, A. J., Zoller, M. & Smith, M. (1982). Nature, 299, 756–758. Google Scholar
Yabukarski, F., Doukov, T., Mokhtari, D. A., Du, S. & Herschlag, D. (2022). Acta Cryst. D78, 945–963. Web of Science CrossRef IUCr Journals Google Scholar
Yabukarski, F., Doukov, T., Pinney, M. M., Biel, J. T., Fraser, J. S. & Herschlag, D. (2022). Sci. Adv. 8, eabn7738. Web of Science CrossRef PubMed Google Scholar
Yuno, A., Lee, M. J., Lee, S., Tomita, Y., Rekhtman, D., Moore, B. & Trepel, J. B. (2018). Methods Mol. Biol. 1709, 423–441. Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.
