Abstract

This column highlights recently published articles that are of interest to the readership of this publication. We encourage ABRF members to forward information on articles they feel are important and useful to Clive Slaughter, AU-UGA Medical Partnership, 1425 Prince Avenue, Athens GA 30606. Tel; (706) 713-2216: Fax; (706) 713-2221: Email; [email protected] or to any member of the editorial board. Article summaries reflect the reviewer’s opinions and not necessarily those of the Association.

NUCLEIC ACID SEQUENCING AND GENOTYPING

Arslan S, Garcia F J, Guo M, Kellinger M W, Kruglyak S, LeVieux J A, Mah A H, Wang H, Zhao J, Zhou C, Altomare A, Bailey J, Byrne M B, Chang C, Chen S X, Cho B, Dennler C N, Dien V T, Fuller D, Kelley R, Khandan O, Klein M G, Kim M, Lajoie B R, Lin B, Liu Y, Lopez T, Mains P T, Price A D, Robertson S R, Taylor-Weiner H, Tippana R, Tomaney A B, Zhang S, Abtahi M, Ambroso M R, Bajari R, Bellizzi A M, Benitez C B, Berard D R, Berti L, Blease K N, Blum A P, Boddicker A M, Bondar L, Brown C, Bui C A, Calleja-Aguirre J, Cappa K, Chan J, Chang V W, Charov K, Chen X, Constandse R M, Damron W, Dawood M, DeBuono N, Dimalanta J D, Edoli L, Elango K, Faustino N, Feng C, Ferrari M, Frankie K, Fries A, Galloway A, Gavrila V, Gemmen G J, Ghadiali J, Ghorbani A, Goddard L A, Guetter A R, Hendricks G L, Hentschel J, Honigfort D J, Hsieh Y-T, Hwang Fu Y-H, Im S K, Jin C, Kabu S, Kincade D E, Levy S, Li Y, Liang V K, Light W H, Lipsher J B, Liu T-l, Long G, Ma R, Mailloux J M, Mandla K A, Martinez A R, Mass M, McKean D T, Meron M, Miller E A, Moh C S, Moore R K, Moreno J, Neysmith J M, Niman C S, Nunez J M, Ojeda M T, Ortiz S E, Owens J, Piland G, Proctor D J, Purba J B, Ray M, Rong D, Saade V M, Saha S, Tomas G S, Scheidler N, Sirajudeen L H, Snow S, Stengel G, Stinson R, Stone M J, Sundseth K J, Thai E, Thompson C J, Tjioe M, Trejo C L, Trieger G, Truong D N, Tse B, Voiles B, Vuong H, Wong J C, Wu C-T, Yu H, Yu Y, Yu M, Zhang X, Zhao D, Zheng G, He M, Previte M. Sequencing by avidity enables high accuracy with low reagent consumption. Nature Biotechnology 42;2024:132-138.

Arslan et al. document sequencing methodology developed by Element Biosciences, Inc., San Diego, CA, for use in its commercial AVITI instrument system. The methodology represents a new implementation of sequencing-by-synthesis. After hybridizing a library of DNA fragments to an instrument flow cell for sequencing, fragment amplification is performed by rolling circle amplification instead of the more usual bridge amplification process. The authors refer to the resulting clonal products as polonies (terminology borrowed from George Church and collaborators) rather than clusters. In the presence of an engineered polymerase, successive nucleotides in the templates are then identified by cycles of hybridization to dye-labeled nucleotides that incorporate a 3′ blocking group. In each cycle, the bound nucleotides are identified by their fluorescence, and then removed and replaced with an unlabeled, blocked nucleotide for chain extension. These nucleotides are finally deblocked in readiness for repetition of the cycle. The dye-labeled nucleotides, however, are supplied as multivalent species termed avidites. Avidites are comprised of a dye-conjugated streptavidin core to which are bound nucleotides that are linked to conjoined biotin molecules to form particles capable of interacting with multiple DNA templates within polonies. The multivalent binding enhances the specificity of interaction, and permits cost reduction by minimizing the concentration of avidites required for the process. The authors report high base-calling accuracy, averaging one error in 10,000 base-pairs (Q40). Notably, accuracy is maintained following long homopolymer sequences. The methodology is compatible with diverse sequencing applications, including single-cell RNA-seq, whole genome sequencing, and single-nucleotide variant identification and indel detection.

Cao L, Kong Y, Fan Y, Ni M, Tourancheau A, Ksiezarek M, Mead E A, Koo T, Gitman M, Zhang X-S, Fang G. mEnrich-seq: methylation-guided enrichment sequencing of bacterial taxa of interest from microbiome. Nature Methods 21;2024:236-246.

Microbiomes usually contain very large numbers of microbial species, many of which may not be amenable to culturing. Microbiome samples also contain host cells. Conventional metagenomic sequencing results in sampling of cells according to their abundance, making the rarest ones – often the most interesting – challenging to study. Cao et al. seek to overcome this inherent bias by implementing methodology favorable for the study of less abundant species. They make use of the observation that patterns of DNA methylation in bacteria are highly sequence-specific and are stable within strains/species, yet strongly divergent between strains/species. They utilize these patterns to enrich bacterial strains of interest within microbiome samples prior to sequencing. They first shear DNA to a size ∼10 kb, and ligate to barcode adaptors. They then select sequences that are methylated in a bacterial taxon of interest, and choose restriction enzymes that cleave sites within these regions in non-methylated sequences in order to degrade genomic DNA from other taxa (and host DNA). Following size selection to recover surviving, longer DNA fragments, they perform PCR amplification, and perform nanopore sequencing. In this way, they enrich targeted taxa up to 117x from human fecal and urine samples, although they comment that even 5x enrichment would be strongly advantageous. Some 4,600 bacterial methylomes have already been mapped, and the authors estimate that 68% of these can be targeted by at least one restriction enzyme for implementation of their method. They comment that many more have yet to be mapped, concluding that the methodology has even broader scope for simplifying the study of complex microbiomes.

Emani P S, Geradi M N, Gürsoy G, Grasty M R, Miranker A, Gerstein M B. Assessing and mitigating privacy risks of sparse, noisy genotypes by local alignment to haplotype databases. Genome Research 33;2023:2156-2173.

Large-scale projects such as the U.K. Biobank and the National Institute of Health’s All of Us, house genetic data intended for discovery of personalized clinical interventions. But those data, if misused or stolen, may also jeopardize privacy. Using a hidden Markov model approach, Emani et al. here provide computational tools for assessing the risk that released sets of SNPs could identify individuals from particular environmental samples. Using a database with ∼5000 haplotypes, they find that as few as 10 common, noise-free SNPs are sufficient to identify individuals, ∼20 SNPs can identify both individuals in a 2-individual composite sample, and 20-30 SNPs can identify first-order relatives. With noisy environmental samples, ∼30 SNPs can identify an individual. Similar risks would be incurred if investigators were to release datasets in the interests of scientific transparency. The authors provide a routine for selective removal of SNPs from datasets that would effectively preserve individuals’ identity. They also note that their methodology may be useful in non-forensic contexts, such as the mapping of somatic mutations in cells to cancer lineages, or viral mutations to clades.

PROTEIN CHARACTERIZATION

McMullan G, Naydenova K, Mihaylov D, Yamashita K, Peet M J, Wilson H, Dickerson J L, Chen S, Cannone G, Lee Y, Hutchings K A, Gittins O, Sobhy M A, Wells T, El-Gomati M M, Dalby J, Meffert M, Schulze-Briese C, Henderson R, Russo C J. Structure determination by cryoEM at 100 keV. Proceedings of the National Academy of Sciences 120;2023:e2312905120.

Electron cryomicroscopy (cryoEM) has proven highly effective in solving the atomic structures of biological macromolecules, but the exceedingly high costs of acquisition, installation and operation of state-of-the-art systems for this purpose restrict availability of the methodology to a small number of national centers. McMullen et al. here describe a prototype instrument built for one-tenth the cost that incorporates an electron source operating at an accelerating voltage of 100 keV, rather than 300 keV used by state-of-the-art systems. The lower energy also obviates the need for the greenhouse gas sulfur hexafluoride to prevent arcing, which further reduces cost, complexity and environmental impact of the system. The authors determine the structures of 11 proteins with their system, achieving resolutions of 2.6-4.5 Å – all sufficient to build accurate atomic models, although not as good as the best recorded cryoEM systems, which are capable of 1.2 Å resolution. They collect images manually at a rate of up to 60 images per hour, requiring a few hours to collect images for each specimen. Although this is much slower than automated microscopes, the system affords convenience in screening specimens for suitability, which is generally a major limiting factor in throughput. It is hoped that further development of this instrumentation will greatly expand access to cryoEM.

Terwilliger T C, Liebschner D, Croll T I, Williams C J, McCoy A J, Poon B K, Afonine P V, Oeffner R D, Richardson J S, Read R J, Adams P D. AlphaFold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination. Nature Methods 21;2024:110-116.

Models of protein structure derived from methodologies such as AlphaFold and RoseTTAFold are known for their unprecedented accuracy. Terwilliger et al. survey the current accuracy level of protein structural predictions provided by AlphaFold by comparing the predicted structures with structural models derived experimentally by X-ray crystallography. In a set of 102 X-ray crystallographic structures selected to bear free R factor values of 0.3 or better, they conclude that AlphaFold predictions often provide strong hypotheses for local structural features and mechanisms of action. AlphaFold enables experiments with specific predictions to be designed for hypotheses to be tested. However, the authors find many instances of AlphaFold predictions deviating substantially either in global domain orientations or locally in Cα backbone conformation or side-chain orientation. The authors do emphasize that both predicted models and crystallographic structures have limitations. A single-crystal electron density map merely provides a single representation of what may best be regarded as a structurally dynamic system, putatively affected by temperature, electrolytes, ligands and macromolecular interactions. AlphaFold models disregard factors that include covalent modifications, ligands, and macromolecular interaction. The study underscores the need for experimental structure determination to confirm AlphaFold structure predictions, especially when molecular interactions not included in the predictions are at issue.

PROTEOMICS

Madern M, Reiter W, Stanek F, Hartl N, Mechtler K, Hartl M. A causal model of ion interference enables assessment and correction of ratio compression in multiplex proteomics. Mol Cell Proteomics 23;2024:100694.

The multiplexing enabled by isobaric labeling systems such as tandem mass tags (TMT) provides enhanced sample throughput, elimination of missing values, and reduction in sample-specific technical variation. However, when peptides from differentially expressed proteins (“target peptides”) are examined, if more than one precursor ion is isolated for fragmentation within the same m/z window, and the co-isolated peptides are derived from proteins that are not expressed differentially, the ratio of intensities of the reporter ions derived from the target peptide may be smaller than their true abundance ratio. This phenomenon is termed ratio compression. The ion interference that causes ratio compression is freshly examined in this paper. The authors measure ion interference in product ion spectra using a two-proteome experimental system with known ground-truth. They conclude that precursor ion purity as quantified by existing metrics does not explain the level of reporter ion interference. Instead, it is caused principally by ions hidden within the noise of the precursor spectra, presumed to be due to other labeled peptides of low abundance. Based on this insight, the authors develop a multiple linear regression model that accurately predicts the level of interference in the reporter ion signal in product ion scans. Because the model is trained directly on each new dataset, it permits correction of ratio compression applicable to any TMT experiment. Software for this purpose is made freely available.

FUNCTIONAL GENOMICS & PROTEOMICS

Russell A J C, Weir J A, Nadaf N M, Shabet M, Kumar V, Kambhampati S, Raichur R, Marrero G J, Liu S, Balderrama K S, Vanderburg C R, Shanmugam V, Tian L, Iorgulescu J B, Yoon C H, Wu C J, Macosko E Z, Chen F. Slide-tags enables single-nucleus barcoding for multimodal spatial genomics. Nature 625;2024:101-109.

Russell et al. explain in proof-of-principal studies how high-throughput analysis of gene expression and DNA accessibility in cell nuclei isolated from tissue sections in a spatially defined manner can provide insight into cellular interactions in histology and pathology. The methodology, termed Slide-tag, uses spatially indexed, monolayer arrays of densely packed, DNA-barcoded, 10-µm beads. A 20-µm frozen tissue section is applied to such an array, and the DNA barcodes are allowed to diffuse into the tissue after photocleaving them from the beads. In the tissue, they associate with, and tag, cell nuclei. Nuclei are then separated in bulk and subjected to single-nucleus sequencing using droplet-based microfluidic methods. Russell et al. use this methodology to assign spatial locations to the diverse cell-types comprising the central nervous system in mouse hippocampus and human cerebral cortex. They also study interactions between subpopulations of B and T lymphocytes in human tonsils. In metastatic melanoma, they document processes occurring in tumor evolution by performing simultaneous analysis of transcriptome and epigenome of neoplastic and stromal cells, and the T-cell receptor repertoire of infiltrating immune cells. The results indicate the value of the methodology for discovery of functional interactions between normal and malignant cells in tissues.

Mason J W, Chow Y T, Hudson L, Tutter A, Michaud G, Westphal M V, Shu W, Ma X, Tan Z Y, Coley C W, Clemons P A, Bonazzi S, Berst F, Briner K, Liu S, Zécri F J, Schreiber S L. DNA-encoded library-enabled discovery of proximity-inducing small molecules. Nature Chemical Biology 20;2024:170-179.

There is potential therapeutic benefit to be derived from small molecules that induce proximity between a protein drug target and a second, endogenous protein that can modulate the target’s activity when the two proteins bind together. Mason et al. seek to formulate a general method to screen for chemical inducers of proximity as potential therapeutic agents. As a test system, they employ protein bromodomains as targets of interest, and von Hippel-Lindau protein (VHL) as the modulator. VHL is an E3 ligase, which exerts its modulating effect by ubiquitinating the target, leading to its degradation by the proteasome. They identify small molecules best capable of inducing proximity between the bromodomain BRD4 and VHL in a library of bifunctional compounds, the members of which bear DNA barcodes. To design a suitable compound library, they begin with a known BRD4 ligand, MZ1. MZ1 has a thiazole methyl group that is suitable for DNA attachment (via an alkyne group for copper(I)-catalyzed azide-alkyne cycloaddition to the DNA headpiece) without interfering with binding to either protein. DNA barcode sequences are constructed by a split-and-pool synthesis protocol. They then attach to MZ1 a VHL ligand via a set of 15 alternative connectors chosen empirically from a panel of 22 structures to minimize disruption of binding to the two proteins. The connectors bear a primary or secondary amine, which the authors couple to a triazine moiety. Triazine is a heterocyclic species that bears two remaining amines that can be functionalized by successive nucleophilic aromatic substitutions. With the 15 connectors, 290 amines for the first of these triazine substitutions and 249 for the second of these triazine substitutions, the authors generate a library of over a million potential test compounds. An in vitro affinity screen for binding to both BRD4 and VHL is then conducted, and the strongly binding compounds are identified by high-throughput DNA sequencing. Their ability to effect bromodomain degradation is then demonstrated in cells. The generality of the method is demonstrated by screening BRD2 and BRDT as additional targets of interest. This methodology is hoped to enable the degradation of hitherto undruggable targets.

IMAGING

Voigt F F, Reuss A M, Naert T, Hildebrand S, Schaettin M, Hotz A L, Whitehead L, Bahl A, Neuhauss S C F, Roebroeck A, Stoeckli E T, Lienkamp S S, Aguzzi A, Helmchen F. Reflective multi-immersion microscope objectives inspired by the Schmidt telescope. Nature Biotechnology 42;2024:65-71.

Yu C-H, Yu Y, Adsit L M, Chang J T, Barchini J, Moberly A H, Benisty H, Kim J, Young B K, Heng K, Farinella D M, Leikvoll A, Pavan R, Vistein R, Nanfito B R, Hildebrand D G C, Otero-Coronel S, Vaziri A, Goldberg J L, Ricci A J, Fitzpatrick D, Cardin J A, Higley M J, Smith G B, Kara P, Nielsen K J, Smith I T, Smith S L. The Cousa objective: a long-working distance air objective for multiphoton imaging in vivo. Nature Methods 21;2024:132-141.

Two groups design novel microscope objectives with a large field of view and a long working distance to improve methods for high-resolution imaging of large samples subjected to tissue clearing procedures, such as entire mouse brains. Such applications require a high numerical aperture, a long working distance, and a wide field of view. Voigt et al. note that different clearing methods require the use of diverse immersion media with a broad range of refractive indices (n) extending from water with n = 1.33 for expansion microscopy to organic solvents such as dibenzyl ether or ethyl cinnamate with n = 1.56 for other applications. They describe a single, simple design concept compatible with this entire range of immersion media. The design uses a mirror instead of lenses. The advantage of a mirror is that reflection by a mirror is independent of the refractive index of the medium in contact with the mirror, so monochromatic aberrations stay the same whatever the refractive index of the medium. Moreover, mirrors provide a longer working distance than lenses of similar numerical aperture. The authors’ objective design emulates the Schmidt wide-field telescope often used in astronomy. It has a spherical mirror, and a correction plate that corrects spherical aberrations from the mirror. In the present application, the mirror and correction plate are immersed in an exchangeable immersion medium, and the sample is placed between them. The surface of the correction plate is deformed so that rays emerging from it on their way to the mirror emerge at normal incidence. This precludes refraction at the interface between the correction plate and the medium, which would otherwise cause aberration when using a different medium. The authors demonstrate the new objective in multi-photon applications. They achieve a numerical aperture of 1.08 at n = 1.56, a 1.1-mm field of view and an 11-mm working distance. In the present implementation, the positioning of the sample in the excitation path does result in diffraction artifacts with large samples, but the authors find that samples up to 5 mm in diameter are accommodated without serious loss of image quality. A wide variety of immersion media are employed successfully with the design, although, unfortunately, good image quality is not achieved with CLARITY samples because the required refractive index matching solutions are not homogeneous enough. A patent application has been filed for this objective design. Yu et al. tackle similar issues with a new lens-based objective. It is designed for convenient compatibility with commonly used multiphoton imaging systems, and uses air as the immersion medium. It provides a long working distance of 20 mm, and a large field of view of 2 mm, but affords a smaller numerical aperture of 0.50. The lens description is open source, and will not be patented.

CELL BIOLOGY

Huang X, Henck J, Qiu C, Sreenivasan V K A, Balachandran S, Amarie O V, Hrabě de Angelis M, Behncke R Y, Chan W-L, Despang A, Dickel D E, Duran M, Feuchtinger A, Fuchs H, Gailus-Durner V, Haag N, Hägerling R, Hansmeier N, Hennig F, Marshall C, Rajderkar S, Ringel A, Robson M, Saunders L M, da Silva-Buttkus P, Spielmann N, Srivatsan S R, Ulferts S, Wittler L, Zhu Y, Kalscheuer V M, Ibrahim D M, Kurth I, Kornak U, Visel A, Pennacchio L A, Beier D R, Trapnell C, Cao J, Shendure J, Spielmann M. Single-cell, whole-embryo phenotyping of mammalian developmental disorders. Nature 623;2023:772-781.

Saunders L M, Srivatsan S R, Duran M, Dorrity M W, Ewing B, Linbo T H, Shendure J, Raible D W, Moens C B, Kimelman D, Trapnell C. Embryo-scale reverse genetics at single-cell resolution. Nature 623;2023:782-791.

In this pair of papers, the authors show how indexing-based, single-cell RNA sequencing (scRNA-seq) of whole embryos provides a scalable, systematic approach to comprehensive understanding at the molecular and cellular levels of the ways in which specific mutations perturb normal developmental processes to produce developmental disorders. As models of human developmental disorders, Huang et al. utilize the mouse, while Saunders et al. utilize the zebrafish. Both studies employ a reverse genetic approach in which the effects on development of specific mutations are investigated. Both studies are conducted at large scale. Huang et al. study 22 different mutants in 103 mouse embryos derived from 4 wild-type strains (generally, 4 replicates per strain), and profile a total of 1.6 million nuclei. However, they restrict attention to a single developmental stage, embryonic day (E)13.5. Saunders et al. study 23 different genes (≥16 replicates per genotype) in 1,812 zebrafish embryos that encompass 19 different developmental time-points, and profile a total of 3.2 million cells. The studies seek to identify perturbations in the composition of embryos with respect to cell type, differences in gene expression within cell types, and shared phenotypic features of different genotypes, to establish how specific or non-specific the perturbations caused by different genotypes might be, and how much pleiotropy exists. Important facets of these studies include four particular points. Firstly, numbers of replicates affect the sensitivity with which changes may be detected against the background of individual variation: large changes require fewer replicates, and more replicates are needed to detect changes in rarer cell types. Secondly, although only a small fraction of the cells in any particular embryo may be profiled, cells from embryos of other genotypes in the dataset may be used to help detect mutant-specific phenotypes of even rare cell types. Thirdly, affects not identified in the present studies may come to light when results from other studies are compiled into a common database, once robust validation strategies are developed for the purpose. Finally, Saunders et al. comment that extension of the methodology to capture both single-cell transcriptomes and the presence of perturbed alleles would help detect instances of mosaicism.

Huppertz M-C, Wilhelm J, Grenier V, Schneider M W, Falt T, Porzberg N, Hausmann D, Hoffmann D C, Hai L, Tarnawski M, Pino G, Slanchev K, Kolb I, Acuna C, Fenk L M, Baier H, Hiblot J, Johnsson K. Recording physiological history of cells with chemical labeling. Science 383;2024:890-897.

A recorder is a molecular device for marking cells when they activate a transient physiological process. Recorders enable the history, and geography, of cells involved in the process to be documented. Huppertz et al. describe a new recorder construct for marking cells in tissues, and provides permanent marks that can be employed subsequently for identification or sorting of cells for further analysis. It also provides the capability of adding distinguishable marks during successive periods in which the physiologic process may be active in the cell population. The construct is derived from a modified haloalkane dehydrogenase that binds to synthetic, ligands, termed HaloTags. HaloTags consist of a chloroalkane linker attached to any of a variety of fluorescent dyes. In this study, the dehydrogenase protein is circularized by connecting the N-terminus to the C-terminus. New N- and C-termini are then created by removing a 10-residue peptide from within the sequence in the region of the ligand-binding site, thereby inactivating the enzyme, but otherwise preserving its 3-D structure. When this split protein, named cpHalo∆, is supplied with an analog of the missing peptide (Hpep) and is brought sterically close to the peptide, the Hpep binds to cpHalo∆, the active dehydrogenase is reconstituted, and rapidly acquires a fluorescent label upon supply of ligand. This system is used to record protein-protein interactions by conjugating cpHalo∆ to a suitable site on one protein and an Hpep to a suitable site on the other protein. Their interaction results in activation and labeling of the recorder construct. The authors extend their application of this system by recording activation of a G protein-coupled receptor, the human dopamine receptor D2. Activation of D2 results in the binding of the cytoplasmic side of the receptor to β-arrestin 2. This interaction can be recorded with HaloTags. The authors also demonstrate the use of a split HaloTag recorder for detection of increases in the intracellular Ca²⁺ concentration. They covalently conjugate cpHalo∆ to an Hpep via a bridge consisting of calmodulin and the peptide M13. Calmodulin binds to M13 in a Ca²⁺-dependent manner, and this binding within the construct produces steric changes that permit interaction between cpHalo∆ and the Hpep. The authors show that when the HaloTag methodology is deployed in cell culture, the recording period is determined by the addition and wash-out of the fluorescent HaloTag ligand: interactions occurring either before or after this interval are not recorded. Temporal resolution in cell culture is therefore determined by the permeability of the fluorescent tags. In animal experiments, temporal resolution is determined by the pharmacokinetics of the tags. The fluorescent signal persists for days, thereby enabling downstream analyses, including cell sorting and transcriptomics. Successive activations may be recorded by supplying different HaloTag fluorophores. The authors anticipate that this system will prove useful in applications such as tracking neuronal activities in flies and zebrafish.

DESIGN & DEVELOPMENT OF THERAPEUTICS

Weng C, Faure A J, Escobedo A, Lehner B. The energetic and allosteric landscape for KRAS inhibition. Nature 626;2024:643-652.

The GTPase KRAS is an oncoprotein constitutively activated by mutation in a large number of cancers. Its GTP-bound state, but not its GDP-bound state, binds a variety of effector proteins to stimulate cell proliferation. Control of these protein interactions is exerted by allosteric changes in KRAS structure upon nucleotide binding. Cancer driver mutations of KRAS stabilize the GTP-bound state, and activate KRAS constitutively. Despite the important role of KRAS in malignancy, KRAS was long considered undruggable. But recently, allosteric inhibitors that bind outside the KRAS nucleotide and effector binding sites and stabilize its GDP-bound state have been approved for clinical use. Weng et al. here reveal a comprehensive map of allosteric sites on KRAS to guide future drug development. They quantify the effects of every possible single amino acid substitution in the protein (>3,200 in total), and >23,300 double substitutions, with respect to binding with 6 different interaction partners. Binding is quantified using a previously published protein-fragment complementation assay in which the binding partners are fused with different fragments of a reporter enzyme, dihydrofolate reductase (DHFR). Their interaction brings the two enzyme fragments into proximity, reconstituting DHFR activity, which promotes cell growth. A second assay is used to quantify the cellular abundance of the KRAS variants in the cells. In the latter assay, only one of the DHFR fragments is fused to the KRAS variant: the other fragment is over-expressed in unconjugated form. Here, growth is proportional to the abundance of the KRAS variant. From these data, binding energies between KRAS variants and effector proteins can be estimated. The authors estimate >22,000 free energy values in this way, and thence map landscapes of sites for putative inhibition of allostery. The methodology is expected to be applicable to mapping allosteric sites on many other proteins, thus rendering further previously undruggable proteins amenable to therapeutic targeting.