{"paper_id":"0ad8a78a-4287-4ede-a6c1-7e42cf8daf97","body_text":"1 \nSecondary structure determines electron transport in peptides 1 \n 2 \nRajarshi Samajdar1,2†, Moeen Meigooni2,3†, Hao Yang2,4, Jialing Li1,2, Xiaolin Liu5, 3 \nNicholas E. Jackson2,5, Martín A. Mosquera6, Emad Tajkhorshid2,3,5,7,8, Charles M. 4 \nSchroeder1,2,3,4,5,8* 5 \n 6 \n1Department of Chemical and Biomolecular Engineering, University of Illinois at Urbana-7 \nChampaign, Urbana, Illinois, 61801, United States  8 \n2Beckman Institute for Advanced Science and Technology, University of Illinois at 9 \nUrbana-Champaign, Urbana, Illinois, 61801, United States 10 \n3Center for Biophysics and Quantitative Biology, University of Illinois at Urbana-11 \nChampaign, Urbana, IL 61801, United States 12 \n4Department of Materials Science and Engineering, University of Illinois at Urbana-13 \nChampaign, Urbana, Illinois, 61801, United States 14 \n5Department of Chemistry, University of Illinois at Urbana-Champaign, Urbana, Illinois, 15 \n61801, United States 16 \n6Department of Chemistry and Biochemistry, Montana State University, Bozeman, MT 17 \n59717, United States 18 \n7Department of Biochemistry, University of Illinois at Urbana-Champaign, Urbana, 19 \nIllinois, 61801, United States 20 \n8Department of Bioengineering, University of Illinois at Urbana-Champaign, Urbana, 21 \nIllinois, 61801, United States 22 \n 23 \n†These authors contributed equally 24 \n*Corresponding author  25 \n 26 \n 27 \n 28 \n 29 \n 30 \n 31 \n 32 \n 33 \n 34 \n 35 \n 36 \n 37 \n 38 \n 39 \n 40 \n 41 \n 42 \n 43 \n 44 \n 45 \n 46 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 2 \nAbstract 47 \nProteins play a key role in biological electron transport, but the structure-function 48 \nrelationships governing the electronic properties of peptides are not fully understood. 49 \nDespite recent progress, understanding the link between peptide conformational 50 \nflexibility, hierarchical structures, and electron transport pathways has been challenging. 51 \nHere, we use single-molecule experiments, molecular dynamics (MD) simulations, non-52 \nequilibrium Green’s function -density functional theory (NEGF-DFT) calculations, and 53 \nunsupervised machine learning to understand the role of primary amino acid sequence 54 \nand secondary structure on charge transport in peptides . Our results reveal a two-s tate 55 \nmolecular conductance behavior for peptides across several different amino acid 56 \nsequences. MD simulations and Gaussian mixture modeling are used to show that this 57 \ntwo-state molecular conductance behavior arises due to the conformational flexibility of 58 \npeptide backbones, with a high-conductance state arising due to a more defined 59 \nsecondary structure (beta turn) and a low-conductance state occurring for extended 60 \npeptide structures. Conformer selection for the peptide structures is rationalized using 61 \nprincipal component analysis (PCA) of intramolecular hydrogen bonding distances 62 \nalong peptide backbones. Molecular conformations from MD simulations are used to 63 \nmodel charge transport in NEGF-DFT calculations, and the results are in reasonably 64 \ngood agreement with experiments. Projected density of states (PDOS) calculations and 65 \nmolecular orbital visualizations are further used to understand the role of amino acid 66 \nside chains on transport. Overall, our results show that secondary structure plays a key 67 \nrole in electron transport in peptides, which provides new avenues for understanding the 68 \nelectronic properties of longer peptides or proteins. 69 \n 70 \n 71 \nSignificance Statement 72 \nElectron transport in proteins serves as a biological power line that fuels cellular 73 \nactivities such as respiration and photosynthesis. Within cells, proteins act as conduits, 74 \nshuttling electrons through a series of reactions and pathways to generate proton 75 \ngradients and to fuel ATP synthesis. Despite recent progress, the mechanisms 76 \nunderlying the flow of energy in protein complexes are not fully understood. Here, we 77 \nstudy electron transport in peptides at the single-molecule level by combining 78 \nexperiments and molecular modeling. Our results reveal two distinct molecular sub-79 \npopulations underlying electron transport that arise due to the flexib ility of peptide 80 \nbackbones and the ability to fold into compact structures. This  work provides a basis for 81 \nunderstanding energy flow in larger proteins or biomolecular assemblies. 82 \n  83 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 3 \nIntroduction  84 \nElectron transport in proteins is essential for maintaining fundamental life processes 85 \nsuch as respiration  and photosynthesis 1. In recent years, a wide range of experiments 86 \nand theoretical studies has focused on understanding electron transfer in biological 87 \nsystems 2–4, ranging from redox events in metalloproteins  5,6 and redox-active cofactors  88 \n7,8 to metal-reducing bacteria  9 . Recent work has shown that proteinaceous nanowire 89 \nfilaments of metal-reducing bacteria such as Geobacter sulfurreducens  exhibit 90 \nremarkable abilities for long-distance electron transport on the micron scale 10,11. During 91 \nsuch redox-mediated electron transport events, intervening residues between redox 92 \ncenters are thought to provide a conductive matrix for electron transport  12. However, 93 \nproteins exhibit complex secondary structures due to intramolecular hydrogen (H)-94 \nbonding interactions within the underlying conductive protein matrix . Despite recent 95 \nprogress, understanding how secondary structure formation in peptides and proteins 96 \naffects electron transport is not yet fully understood. 97 \n 98 \nElectron transport in molecules can occur by different mechanisms such as single-step 99 \n(coherent) tunneling, multi-step (incoherent) hopping, resonant tunneling, or flickering 100 \nresonant tunneling 13–15. The dominant mechanism for nanoscale charge transport in 101 \nshort peptide sequences has been reported as non-resonant coherent tunneling 3,4,16–22, 102 \nwhere conductance decays exponentially with molecular length. However, electron 103 \ntransport in long peptide or protein sequences also occurs  by hopping 7,23, where 104 \nconductance decreases  inversely with distance. The environment around a protein 105 \naffects the driving force for the electron transfer reaction and the reorganization energy, 106 \nin accordance with Marcus theory 24. Molecular conformation and intramolecular H-107 \nbonding that arise due to the protein sequence and environment are pivotal for 108 \ncontrolling biological electron transport over long distances 25. Prior work has focused on 109 \nunderstanding electron transport in helical peptides 26–29 using bulk conductivity , 110 \nelectrochemistry, thin-film conductivity, or electronic measurements on assembled 111 \npeptide monolayers 7,19,27. However, key knowledge gaps remain in understanding how 112 \nother types of secondary structures in peptides and proteins affect electron transport in 113 \nbiological systems. Elucidating electron transport at the single-molecule level holds the 114 \npotential to provide valuable new insights into the electronic properties of more complex 115 \npeptide or protein structures.  116 \n 117 \nSingle-molecule techniques offer the ability to characterize conformation-dependent 118 \nelectron transport in the absence of intermolecular interactions in monolayers  or bulk-119 \nscale measurements. In recent years, single-molecule conductance measurements for 120 \npeptides have primarily focused on short peptide sequences containing up to two or 121 \nthree amino acids 22,30 or chemically functionalized peptides  to facilitate metal electrode 122 \ncontact31. However, peptide backbones are generally more flexible compared to -123 \nconjugated carbon backbones commonly used in synthetic organic electronic materials, 124 \nand this enhanced backbone flexibility could give rise to conformation-dependent 125 \nelectron transport pathways in oligopeptides. By using the scanning tunneling 126 \nmicroscope break junction (STM- BJ) technique, the phenomenon of electron tunneling 127 \nwhile pulling 32,33 single  molecules  has been studied . In addition, it has been reported 128 \nthat a special arrangement of hydrogen bonds 34,35 could give rise to conducting 129 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 4 \npathways in non-conjugated peptide backbones. From this view, single-molecule 130 \nmethods offer intriguing routes to understand the role of amino acid sequence and the 131 \neffect of secondary structure on molecular charge transport in peptides, thereby adding 132 \nnew insights into electronic phenomena and their correlation to structure in biomolecular 133 \nsystems. 134 \n 135 \nThe ability to combine molecular simulation with single-molecule electronics 136 \nexperiments provides a powerful approach to understand biophysical processes. The 137 \nrich conformational space of biomolecules 36 such as peptides 37 can be explored using 138 \nmolecular dynamics (MD) simulations. Biomolecular simulation offers a predictive tool 139 \nfor structural biology due to the high spatial and temporal resolutions and the 140 \nextensively tested and validated force fields 38,39. MD simulations have been used to 141 \nunderstand the influence of atomic structure on the electronic properties of synthetic 142 \norganic materials by modeling the structural dynamics of molecular junctions  40–42. 143 \nHowever, classical force fields are limited in their description of molecular junctions that 144 \ninvolve transition metal atoms such as gold. Incorporating Au atoms into classical MD 145 \nsimulations requires either a physical ly rigorous but computationally demanding 146 \nquantum mechanical (QM) description of gold and its interaction with the surrounding 147 \nsystem, or an approximate but more computationally feasible model of interactions with 148 \nAu atoms. Examples of the latter include representing gold atoms as dummy particles 149 \nrestricted to only interact with specified anchor atoms through harmonic potentials 41 and 150 \nutilization of reactive force fields to model bond formation and disruption 43. Molecular 151 \nconformations generated by MD can be used in computationally efficient QM 152 \ncalculations for improved comparison between theory and experimental results.  153 \n 154 \nIn this work, we investigate the role of amino acid sequence and secondary structure on 155 \nthe electronic properties of peptides using a combination of experiments and 156 \ncomputational modeling. A key feature of our work lies in using MD simulations to 157 \nunderstand the conformational dynamics of molecular junctions in single-molecule 158 \ncharge transport experiments. A scanning tunneling microscope break junction (STM-159 \nBJ) technique is used to experimentally characterize the molecular charge transport 160 \nproperties of oligopeptide s. Our results reveal a two-state conductance behavior for 161 \npeptide sequences contain ing 4 or 5 amino acids . Our results further indicate that 162 \nlonger amino acid sequences can show enhanced conductance values for the extended 163 \nstate due to the presence of aromatic or constrained amino acid side chains.  Gaussian 164 \nmixture modeling (GMM) and MD simulations are used to show that this two-state 165 \nmolecular conductance behavior arises due to the conformational flexibility of the 166 \npeptide backbone. Classical MD simulations with custom potentials for implicitly 167 \nrepresenting gold are used to understand the molecular basis for conformation-168 \ndependent electron transport in peptides. Characteristic conformers for each peptide 169 \nsequence are selected from MD simulations and quantitatively analyzed using principal 170 \ncomponent analysis (PCA) to understand the role of hydrogen bonding  (H-bonding) 171 \ninteractions along the peptide backbone . Interestingly, results from PCA show that 172 \nspecific H-bonding distances between peptide backbone atoms significantly contribute 173 \nto the  structural variation observed in MD simulations. Molecular conformations from 174 \nMD simulations are then used in non -equilibrium Green’s function -density functional 175 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 5 \ntheory (NEGF -DFT) calculations to understand the role of molecular conformation on 176 \ncharge transport. Projected density of states (PDOS) calculations and molecular orbital 177 \nvisualization are further carried out to understand the role of amino acid side chains and 178 \nthe underlying transport mechanisms. Our results reveal that an extended peptide 179 \nsequence gives rise to a low conductance state, whereas a folded conformation (beta 180 \nturn) gives rise to a high conductance state. Overall, our work highlights the importance 181 \nof molecular conformation and secondary structure on the electron transport behavior of 182 \npeptides.  183 \n 184 \nResults and Discussion 185 \nSingle-molecule conductance measurements and chemical characterization 186 \nTetra- and pentapeptides were designed with different amino acid sequences  to 187 \nunderstand the role of non-polar aliphatic R groups , aromatic R groups, or sterically 188 \nconstrained R groups  on electron transport ( Figures 1a,b,c and Supplementary 189 \nFigures 1- 10). The N- and C-terminal residues of the tetra- and pentapeptides were 190 \nselected as methionine, which contains a methyl sulfide (-S- CH3) group that readily 191 \nbinds to gold 44, thereby providing robust electrical contacts to metal electrodes in STM-192 \nBJ. All STM-BJ measurements on peptides were carried out in water (peptide 193 \nconcentration <1 mM).  194 \n 195 \nCircular dichroism (CD) spectra were first obtained for all tetra- and pentapeptides in 196 \nwater at room temperature under identical solvent conditions used in STM-BJ 197 \nexperiments (Supplementary Figures 11-15). CD spectra clearly indicate the presence 198 \nof H-bonding interactions for all tetra- and pentapeptides and show spectral features 199 \nexpected for 3 10 helices, such as a maximum or minimum around ~200-210 nm and a 200 \nshoulder or small peak around ~220 nm  45,46. CD spectral features for 3 10 helices are 201 \nqualitatively different than the spectral features observed for alpha helices, beta sheets, 202 \nor random coils  47. Based on results from CD experiments, the proline and alanine-203 \nbased peptide sequences show minima in CD spectra around 200 nm, which is 204 \nconsistent with a tendency to adopt right-handed 310 helices. On the other hand, peptide 205 \nsequences containing glycine, tyrosine, and tryptophan show peaks in CD spectra 206 \naround 200 nm, which is consistent with left- handed 310 helices. Overall, these results 207 \nclearly indicate the presence of H-bonding interactions amongst the tetra- and 208 \npentapeptides characterized in single-molecule electronics experiments.  209 \n 210 \nWe began by characteriz ing the electronic properties of peptides containing non-polar 211 \naliphatic R groups. The molecular conductance of oligopeptides was determined using a 212 \ncustom-built STM-BJ instrument ( Figure 1d ), as described in prior work 48,49. Our 213 \nexperiments revealed the presence of two distinct conductance populations, as shown 214 \nin characteristic single-molecule conductance traces ( Figure 1e). We hypothesized that 215 \nthe high and low conductance states  could arise due to a folded , compact conformation 216 \nand an extended peptide conformation, respectively . Characteristic single-molecule 217 \nconductance traces for all tetra- and pentapeptides ( Figures 2a,b) indicate that the two 218 \nconductance states occur in the same individual traces  rather than in two separate 219 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 6 \nmolecular sub-populations. This behavior suggests that dynamic conformational 220 \nchanges during molecular pulling events give rise to multiple conductance states.  221 \n 222 \nOne-dimensional and two-dimensional molecular conductance histograms were 223 \ngenerated for the tetra- and pentapeptides across ensembles of >5000 single molecules 224 \n(Figures 2c,d,e,f  and Supplementary Figures 16-17 ). A bimodal conductance 225 \ndistribution is observed for all oligopeptide sequences across the entire range of applied 226 \nbiases (100 mV - 400 mV) studied in this work ( Supplementary Figure 18 ). Bimodal 227 \nconductance distributions can arise due to conformationally distinct molecular sub-228 \npopulations (static heterogeneity) or due to conformation-dependent conductance 229 \nduring molecular pulling (dynamic heterogeneity). To investigate the origins of this 230 \nbehavior, we determined the most probable conductance of the low and high 231 \nconductance states from a Lorentzian fit to the conductance data 50 ( Supplementary 232 \nTables 1-2). The high-conductance peak (~10-2.80 - 10-2.90 G0) occurs at nearly the same 233 \nvalue for all the oligopeptide sequences. The low conductance peak value shows a 234 \nsmall dependence on the backbone sequence and side chain composition. In addition, 235 \nthe molecular displacement corresponding to the low conductance peak is significantly 236 \nlarger than the displacement for the high conductance peak. Based on these results, we 237 \nhypothesized that the low conductance peak arises due to an extended peptide 238 \nconfiguration, whereas the high conductance peak is related to a folded or more 239 \ncompact peptide conformation. 240 \n 241 \nThere are some subtle differences in the low conductance state for the tetra- and 242 \npentapeptides studied in this work , which suggests that amino acid side chain identity 243 \nplays a role in transport . For the tetrapeptides, the low conductance state of MGGM is 244 \n~0.2-0.3 log G0 lower compared to all other sequences (MAAM, MYYM, MWWM, and 245 \nMPPM). These results show that changing the amino acid side chain from hydrogen to 246 \na methyl, aromatic , or a constrained side chain leads to an enhancement in 247 \nconductance. For the pentapeptides, the conductance values for MGGGM and MAAAM 248 \nare approximately half an order- of-magnitude smaller compared to MYYYM, MWWWM, 249 \nand MPPPM. The higher conductance values for MYYYM and MWWWM indicate that 250 \naromatic side chains can lead to enhanced conductance values . MPPPM has a higher 251 \nconductance compared to the glycine or alanine-based sequences, as proline provides 252 \na constrained side chain that reduces the conformational  flexibility and increases the 253 \nrigidity of the molecule . The high er conductance values observed for the extended 254 \nconformations for peptides containing tyrosine, tryptophan, and proline sequences are 255 \nalso corroborated by NEGF-DFT simulations ( Figure 5 e,f), as discussed below. Based 256 \non these results, STM-BJ experiments reveal several intriguing findings regarding the 257 \nrole of amino acid side chains on oligopeptide charge transport.  258 \n 259 \nSingle-molecule data can be quantitatively analyzed using unsupervised learning 260 \nalgorithms to classify molecular charge transport behavior into characteristic groups and 261 \nto identify underlying structure-property relationships 31,51–54. Here, we use silhou ette 262 \nclustering55  (Supplementary Figure 19 ) to determine the optimal number of clusters 263 \nfor data sets corresponding to molecular ensembles for each peptide sequence. 264 \nSilhouette clustering indicates that the optimal number of clusters for all tetra- and 265 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 7 \npentapeptides is two. Gaussian mixture modeling (GMM) is further used to analyze the 266 \ntwo different clusters identified by Silhouette clustering ( Supplementary Figures 20-267 \n21).  268 \n 269 \nResults from GMM show that Cluster 1 accounts for 85-95% of the single-molecule 270 \ntraces and shows both characteristic conductance populations appearing together in the 271 \nsame molecular traces. Cluster 2 accounts for only 5-15% of the data and represents 272 \ntraces in which no molecule is detected or only background signal is observed. If the 273 \nbimodal distribution arose due to stable, conformationally distinct molecular sub-274 \npopulations (static heterogeneity ), then the two characteristic conductance populations 275 \nwould segregate into different clusters. However, our results show that the bimodal 276 \nconductance populations appear sequentially in single-molecule traces for all tetra- and 277 \npentapeptides, which strongly supports conformation-dependent charge transport 278 \nbehavior in peptide backbones (dynamic heterogeneity). 279 \nMD simulations 280 \nTo understand the role of molecular conformation on charge transport, we performed 281 \nMD simulations for all tetra- and pentapeptides ( Figures 1a,b,c) in explicit solvent with 282 \na series of custom potentials to implicitly represent interactions between peptides and 283 \ngold electrodes. These custom potentials and their resulting collective variable 284 \ndistributions are shown in Figures 3a,b,c and Supplementary Figure 22.   285 \n 286 \nThe projection of the end- to-end distance (sulfur anchor- to-anchor distance on terminal 287 \nmethionines) of the peptide along the experimental pulling axis was harmonically 288 \nrestrained to a range of values (6 Å, 9  Å, and 12 Å), allowing the peptide to adopt an 289 \nensemble of conformations. The conformations observed in MD simulations are not 290 \nsignificantly affected by a change in applied voltage ( Supplementary Figure 23), which 291 \nis consistent with single-molecule conductance experiments. Ramachandran free 292 \nenergy plots 56 ( Supplementary Figures 24-25) were determined for the non-terminal 293 \nresidues for all tetra- and pentapeptides. These results indicate that all sequences can 294 \nform left-handed or right-handed helices, except for those  based on proline , consistent 295 \nwith CD measurements.  296 \n 297 \nResults from MD simulations show that backbone hydrogen bonds, which play a key 298 \nrole in defining the secondary structure  of the peptide 57, form with remarkable 299 \nconsistency during the 6 Å end- to-end holding of all peptide sequences considered in 300 \nthis work ( Figures 3d,e  and Supplementary Figures  26-27). However, H-bonding 301 \ninteractions are completely abolished when the end- to-end distance is restrained to a 302 \ndistance of 12 Å. For the  tetra- and pentapeptides considered here, a canonical 303 \nsecondary structure forms  at small end-to-end distances, indicative of a beta turn. A 304 \nbeta turn is defined by an H-bond between the carbonyl oxygen of residue i and the 305 \namide hydrogen of residue i+3 57. In the tetrapeptides, a 1→4 H -bond is observed, 306 \nwhereas for the penta peptides, a 2→5 H -bond is consistently observed. Two 307 \nconformers are selected from the 6 Å and 12 Å holding stages ( Figures 3f,g) of each 308 \npeptide from the peak of the probability distributions of 1→4 H-bond and 2→5 H-bond 309 \ndistances for tetra- and pentapeptides, respectively. It is known that consecutive beta 310 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 8 \nturns in a longer peptide sequence give rise to 3 10 helices57. From this view, our work 311 \nsuggests that helical elements play a key role in the charge transport behavior of 312 \nbiomolecules with defined secondary structures. 313 \n 314 \nWe next performed a linear dimensionality reduction on the MD trajectories to quantify 315 \nhow individual interatomic distances contribute to the peptide conformational landscape. 316 \nThe main objective of this analysis is to identify conserved structural differences across 317 \nall peptides of interest between various end- to-end holding stages ( Figure 4  and 318 \nSupplementary Figures 28-31). Peptides are represented using a Euclidean distance 319 \nmatrix of the common molecular subgraph shared between all sequences. Using this 320 \napproach, each peptide’s structural ensemble is projected onto a shared basis. In 321 \naddition, the i→i+3 H-bonding distances selected as a basis for conformer extraction 322 \nare well captured in the first two principal components, indicating that these distances 323 \ncontribute significantly to the  variance in molecular structure compared to other 324 \ninteratomic distances . Regions of conformational space corresponding to small  i→i+3 325 \ndistances are shown to depopulate with increasing inter-anchor displacement across all 326 \npeptide sequences. Based on these results, MD coupled with unsupervised machine 327 \nlearning (ML)-based data analysis clearly elucidates the key structural features for 328 \ncharacterizing tetra- and pentapeptides in molecular junctions, revealing the most 329 \nprobable peptide conformations. The most probable conformations are then used in 330 \ncomputationally efficient NEGF-DFT calculations to understand the role of molecular 331 \nconformation on the electron transport properties of peptides. 332 \nNEGF-DFT calculations  333 \nTo understand the role of molecular conformation on charge transport in peptide 334 \nbackbones, NEGF-DFT calculations are performed using the most probable simulated 335 \nMD conformations. NEGF-DFT simulations are carried out for extended and turn 336 \nconformations for each peptide sequence ( Figures 5a,b ) using the TranSiesta and 337 \nTbtrans package (Methods). The transmission probabilities as a function of energy 338 \nindicate stark differences between the turn and extended peptide conformations 339 \n(Figures 5c ,d and Supplementary Figures 32-34). The conductance at zero bias 340 \ndiffers significantly between the extended and the turn state of the tetra- and 341 \npentapeptides ( Figures 5e,f ). Our results show reasonable qualitative agreement 342 \nbetween experiments and NEGF-DFT simulations ( Supplementary Table 3-6). Results 343 \nfrom the combined approach of using MD simulations with NEGF-DFT simulations 344 \nsupport the hypothesis that the low conductance population arises from an extended 345 \npeptide conformation, whereas the high conductance population is related to a more 346 \ndefined secondary structure (beta turn) in the peptide. Figures 5e,f also corroborate the 347 \nrole of amino acid side chains  that was observed in experiments on tetra-and 348 \npentapeptides. The glycine-based tetrapeptide sequence has a lower value of the 349 \ntransmission probability near the Fermi level for the extended conformation compared to 350 \nall other sequences. For the pentapeptides, MGGGM and MAAAM show similar 351 \nconductance values in the extended state , albeit lower than MYYYM, MWWWM and 352 \nMPPPM. Overall, NEGF-DFT results qualitatively agree with single-molecule charge 353 \ntransport experiments and provide insights into the role of side  chains on oligopeptide 354 \ncharge transport.  355 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 9 \n 356 \nSite specific PDOS calculations were carried out for all tetra-and pentapeptides 357 \n(Supplementary Figure 35 ) in the extended peptide conformation to understand  the 358 \nrole of amino acid side chain s on electron transport. For all tetrapeptides, PDOS 359 \ncalculations were performed for two carbon atoms along the backbone in the energy 360 \nrange of -5 to 5 eV ( Supplementary Figures 36 a,b). At the Fermi energy level for the 361 \ntetrapeptides (~ -2.20 eV), the PDOS has a relatively low value ( Table 7) due to the 362 \nnon-conjugated peptide backbone. These results imp ly that the backbone orbitals in 363 \npeptides generally yield small er conductance values near the Fermi level compared to 364 \nfully -conjugated systems. To compare these results for the various tetrapeptides, the 365 \nbehavior near the Fermi energy level was investigated ( Supplementary Figures 36 366 \nc,d). Around the Fermi energy level, higher PDOS values are observed for sequences 367 \ncontaining tyrosine and tryptophan. For all pentapeptides, PDOS calculations were also 368 \nperformed for three carbon atoms along the backbone in the energy range of -5 to 5 eV 369 \n(Supplementary Figures 37 a,b,c). The DFT-derived Fermi energy level for the various 370 \npentapeptides is around -2.20 eV, with molecular LUMO levels being above the  Fermi 371 \nlevel by at least 200 meV, though it is important to note that standard DFT generally 372 \nunderestimates the molecular HOMO-LUMO gap. At the Fermi energy level, the value 373 \nof PDOS approach es zero  ( Table 8 ), similar to the case of tetrapeptides . A similar 374 \nanalysis was performed for the pentapeptides near the Fermi energy level 375 \n(Supplementary Figures 37 d,e,f ), showing larger PDOS values for sequences 376 \ncontaining tyrosine and tryptophan for the 1st and 2nd carbon atoms along the backbone. 377 \nFor the 3 rd carbon atom, a relatively high PDOS value is observed for MGGGM , 378 \nMYYYM, and MWWWM. However, the transmission probability for MGGGM is 379 \nsignificantly lower compared to MYYYM and MWWWM . Taken together, these results 380 \nshow that the orbitals of the aromatic side chains tend to mix more readily  with the 381 \nbackbone orbitals compared to other amino acids, which leads to enhancement in 382 \nconductance values.  383 \nPDOS calculations were also carried out for all carbon and hydrogen atoms 384 \n(Supplementary Figure 38) for MGGM, MYYM, MGGGM, and MYYYM in the extended 385 \npeptide conformation . Our results indicate significantly higher PDOS values for the 386 \nsequences containing tyrosine compared to glycine. Overall, these results indicate that 387 \noligopeptide sequences with aromatic side chains have more contribution from the 388 \nbackbone orbitals to the overall electronic density and hence molecular conductance.  389 \nMolecular orbitals were plotted using Siesta 58 and visualized using Vesta 59 390 \n(Supplementary Figures 39- 42). Here, HOMO, HOMO-1, LUMO and LUMO +1 are 391 \nplotted for the glycine and tyrosine-based tetra- and pentapeptides using an isosurface 392 \nvalue of 0.025. These results illustrate relatively weak coupling between the molecules 393 \nand electrodes, which is consistent with the transmission function results  observed for 394 \nthe oligopeptides, in agreement with the proposed tunneling mechanism. These results 395 \nfurther suggest the absence of - stacking interactions between the tyrosine 396 \nsidechains. Overall, these results are consistent with non-resonant tunneling rather than 397 \nresonant tunneling or flickering resonant transport mechanisms for electron transport . 398 \nPrior work by Xiao et al. 22 characterized electron transport in short peptide sequences 399 \nsuch as cysteamine-glycine-glycine -cysteine and cysteine-glycine -cysteine, with results 400 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 10 \nshowing an exponential decay in conductance as a function of molecular length , 401 \nconsistent with single-step tunneling as the dominant transport mechanism.  The tetra-402 \nand pentapeptides based on glycine studied here are of similar length, with the primary 403 \ndifference of methionine as the N- and C-termini amino acids in place of cysteine or 404 \ncysteamine. Based on these results, and the relatively short distance of transport 405 \nobserved in our molecular junctions (< 1.4 nm60), our results are fully consistent with off-406 \nresonant coherent tunneling for the oligopeptides studied in this work.   407 \n 408 \nDuring the STM-BJ pulling experiments, we observe two conductance states related to 409 \ntwo distinct molecular conformations. To further understand the role of H-bonding on 410 \ntransport pathways, we used a bond counting methodology based on the tunneling 411 \npathway model61,62. In general, there is a conductance decay associated with through-412 \nbond, through-space, and through -H-bond electron transport 63. Tunneling is generally 413 \nmore efficient for through-bond compared to through-space transport due to the lower 414 \npotential barrier 64. As a rule of thumb, it can be assumed that the conductance decay 415 \nthrough an H-bond is twice as large compared to the decay through a covalent bond 64. 416 \nSupplementary Figure 43 indicates that if transport were to occur entirely through -417 \nbond, then the pathway would be approximately three bonds longer with an order of 418 \nmagnitude smaller decay compared to the case of electron transport through H-bonds65.  419 \n 420 \nTo further understand the importance of H- bonds on transport, we performed control 421 \nexperiments for STM -BJ using 1,16-hexadecanedithiol ( Supplementary Figure 44) in 422 \n1,2,4-tricholorbenzene. 1,16-hexadecanedithiol has a similar contour length as the 423 \npeptides studied in this work but with a flexible alkane chain backbone and no possibility 424 \nof intramolecular H-bonding. Our results show that two conductance populations are 425 \nobserved for the peptides (at ~10 -2.8 G/G0 and 10 -4.2 G/G0), but no significant 426 \nconductance peaks are observed for the flexible alkane backbones , though a faint 427 \npopulation is observed between ~10-1 -10-2 G/G0, which arises due to the use of different 428 \nanchors and strong binding between the - SH terminal anchor groups  and the gold 429 \nelectrode66. Overall, these results show a two order- of-magnitude increase in molecular 430 \nconductance for a peptide compared to an alkane chain with similar contour  length. We 431 \nfurther compared these results to prior work in the literature. Inkpen et al. 67 studied 432 \ncharge transport in alkane chains such as C 12(SH)2 and C 12(SMe)2, and only a single 433 \nconductance population was observed below ~ 10-5 G/G0 for the C 12 sequences. It 434 \nshould be noted that t he average conductance values reported for the C 12 sequences 435 \nare approximately one  order- of-magnitude lower compared to the low conductance 436 \nstate of the 17- or 19-mer oligopeptide sequences studied in this work . Taken together, 437 \nthese results show that the electron transport behavior of alkane chains is significantly 438 \ndifferent than peptides due to intramolecular backbone H-bonding.  439 \n 440 \nIn this work, we use a combination of single-molecule conductance experiments, MD 441 \nsimulations, and NEGF-DFT calculations to investigate the charge transport properties 442 \nof a series of different peptide sequences. Our results unequivocally reveal the 443 \nstructure-function relationships governing the observed electron transport in peptides, 444 \nhighlighting the importance of secondary structure on charge transport in biomolecules . 445 \nUnsupervised learning is used to analyze single-molecule conductance data, showing 446 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 11 \nthat peptides exhibit a bimodal conductance distribution with a low and high 447 \nconductance population arising from distinct conformational states of peptide 448 \nbackbones. A key feature of our work lies in using MD simulations to sample and 449 \ncharacterize the conformational space of the peptides and to identify conformations to 450 \nbe used in electron transport (NEGF-DFT) calculations . Moving forward, our work could 451 \nprovide new avenues to understand the interplay between molecular charge transport 452 \nand secondary structure in more complex peptide sequences with mixed amino acids 453 \nand/or longer peptides. Proteins are candidate materials for fabricating functional 454 \nmolecular electronic devices due to biocompatibility, anti-fouling properties 68, and 455 \ntunable redox activity due to aromatic amino acids 69. From this view, our work can 456 \nprovide further insights into understanding the role of higher order assembled structures 457 \non biological charge transport, which can be used to inform the design self-assembled 458 \nbioelectronic materials.  459 \n 460 \nMethods 461 \n 462 \nOligopeptide sequences  463 \nAll oligopeptide sequences were purchased from GenScript (Piscataway, NJ). Mass 464 \nspectrometry data for these sequences are provided in the Supplementary Information 465 \n(Supplementary Figures 1-10). 466 \n 467 \nSingle-molecule conductance measurements 468 \nSingle-molecule conductance measurements were performed using a custom-built 469 \nscanning tunneling microscope break junction (STM-BJ) 48,49,66. Gold STM tips were 470 \nprepared using 0.25 mm Au wire (99.998%, Alfa Aesar). STM-BJ experiments were 471 \ncarried out in Milli-Q water (Specific resistance of 18.2 MΩ·cm @ 25 °C ). Due to the 472 \npolarity of the solvent, STM tips were coated with an Apiezon wax to prevent Faradaic 473 \ncurrents from masking characteristic molecular features 70. Gold su bstrates for the 474 \nmeasurements were prepared by evaporating 120 nm of gold on polished AFM metal 475 \ndiscs (Ted Pella). Peptide concentrations  (<1 mM) were selected to yield Poisson 476 \nstatistics in molecular conductance traces. Conductance histograms (> 5000 traces) are 477 \ngenerated for all molecules without data selection. Silhouette clustering and Gaussian 478 \nmixture modelling (GMM) were further used to analyze  the bimodal conductance 479 \ndistribution (Supplementary Information).   480 \n 481 \nMD simulations  482 \nMolecular dynamics (MD) simulations were performed to generate conformational 483 \nensembles for the tetra- and pentapeptide molecular junctions at three anchor 484 \ndisplacements (referred to as stages 6 Å, 9 Å, and 12 Å ). For each peptide, 16 initial 485 \nstructures were prepared using the PeptideBuilder python package 71. Phi and psi 486 \nbackbone dihedrals of each of the 16 structures were randomized independently. Each 487 \nbackbone dihedral angle of non -proline residues  was initialized to a random value 488 \nbetween -180 and 180 degrees, whereas the p hi angle of proline was  initialized to a 489 \nrandom value between -80 and -50 degrees. Hydrogens were added to the peptides 490 \nwith the VMD plugin PSFGEN 72 using the NTER and CTER terminal patches to create 491 \npositively and negatively charged N- and C-termini, respectively. Peptide structures 492 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 12 \nwere then solvated in a cubic box of TIP3P water of side length 38 Å using the VMD 493 \nSOLVATE plugin72. The solvated systems were then subjected to MD simulations with 494 \nthe CHARMM36m protein force field 38,39 using OpenMM 7.7.0 73. Dynamics were 495 \nintegrated using the LangevinMiddleIntegrator 74 with friction coefficient of 1 ps -1, 496 \ntemperature of 300 K, and a timestep of 4 fs. Hydrogen mass repartitioning was not 497 \nutilized. Bonds involving hydrogen atoms, and all bonds and angles involving water 498 \nwere constrained 74. Nonbonded interactions were computed with a cutoff of 12 Å with 499 \nsmooth switching starting at 10 Å. Electrostatic interactions were evaluated using 500 \nparticle mesh Ewald 75 (PME) summation with error tolerance of 0.0005. Each replicate 501 \nwas simulated for 200 ns for each of three holding stages, for a total aggregate 502 \nsimulation time of 96.0 μs (10 peptides × 3 stages × 16 replicates × 200 ns ). The 503 \nconformational ensemble of each peptide is shown to converge after 200 ns of 504 \nsimulation per replicate per holding stage ( Supplementary Figures 45-46). Holding 505 \nstages were  enforced using a series of custom external potentials, applied using 506 \nOpenMM’s custom force classes, described below. The last 190 ns of each simulation 507 \nwas used for subsequent analysis.  508 \n 509 \nA series of custom potentials were implemented to implicitly represent interactions 510 \nbetween the peptide and gold particles. Three potentials were defined: (1) a potential to 511 \nrestrain the distance between the anchors of the molecular junction along the pulling 512 \naxis to 6 Å, 9 Å, or 12 Å (representing the restraints imposed by connections to the gold 513 \nelectrodes); (2) a per-atom charge-dependent potential along the pulling axis 514 \naccounting for electric field forces arising from a voltage-biased junction; and, (3) a 515 \npotential that orients methionine’s thioether moiety such that the average position of 516 \neach sulfur’s lone pairs are oriented towards the (implicitly represented) gold electrodes 517 \nalong the pulling axis. These potentials are described in detail in the next section and 518 \ndepicted in Supplementary Figure 17. 519 \n 520 \nAfter MD simulations, characteristic conformations of each peptide were determined 521 \nfrom their aggregate MD trajectories. For each peptide, two conformations were 522 \nselected from their 6 Å and 12 Å holding-stage simulations at the peaks of their 523 \nrespective hydrogen-bond distance distribution histograms. The H-bond distance 524 \ndistributions used as the basis for conformation selection for the tetra- and 525 \npentapeptides were the 1 ⟶4 and 2 ⟶5 distances, respectively. All free energy plots 526 \n(Figures 4b ,c and Supplementary Figures 21,23 ) were prepared using PyEMMA 527 \n2.5.1176.   528 \n 529 \nCustom potential for implicit gold peptide interactions  530 \nA key  challenge for simulating single-molecule pulling processes is large difference 531 \nbetween the pulling rates used in experiments and those accessible by MD simulations . 532 \nTypical experimental pulling rates are on the order of Angstroms per millisecond (1 Å 533 \nper 5 ms in present study), whereas single-trajectory MD simulations (at most) typically 534 \nreach ms timescales, e.g., with the use of bespoke hardware 77 or massively distributed 535 \ncomputing schemes78. In addition, the need for multiple independent simulation replicas 536 \nto claim ensemble convergence and statistical certainty of key observables further 537 \nrestricts simulations to sub-experimental timescales. However, because the 538 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 13 \nexperimental pulling rate is also slow relative to characteristic relaxation timescales of 539 \nsmall peptides, we assume that all molecular conformations accessible at a given end-540 \nto-end distance are sampled during each step of the experimental pulling process. In 541 \nother words, experimental pulling occurs as an equilibrium process. Rather than 542 \nperforming costly simulations of the entire pulling process, it is more computationally 543 \nfeasible to simulate the molecular junction at various holding (end- to-end distance) 544 \nstages representing the different separation distances arising during the pulling 545 \nexperiments.  546 \n 547 \nUsing this approach, we perform ed a series of independent simulations where we 548 \nrestrained the end- to-end (sulfur-sulfur) distance along the pulling axis to one of three 549 \ndistances spanning the range of end-to-end distances (6 Å, 9 Å, or 12 Å). We define the 550 \npulling axis as the z-axis in our simulations. Schematic illustrations for each potentia l 551 \nare shown in Supplementary Figure 17. The functional form of the potential utilized to 552 \nenforce this restraint is given in Equation 1: 553 \n 554 \n                                                         𝑈1 =\n1\n2 𝑘1[(𝑧𝑆2 − 𝑧𝑠1)−  𝑧0]2, (1) \n 555 \nwhere the coefficient k1 is the force constant of the harmonic potential, zS1 and zS2 are 556 \nthe z-coordinate of the sulfur atoms of the N-terminal and C-terminal methionine 557 \nresidues respectively, and z0 is the equilibrium distance for the given stage. We use a 558 \nvalue of 1 kcal/mol/Å 2 for k1, and we utilize three independent holding stages with z0 559 \nequal to either 6 Å, 9  Å, or 12 Å. This force constant was selected such that the 560 \nresulting distributions of zS2 – zS1 distances have slight overlap ( Supplementary 561 \nFigures 17a,d). 562 \n 563 \nBy restraining the z-displacement between the sulfur atoms, rather than the distance, 564 \nthe movement of each sulfur atom is effectively restrained to one of two parallel planes 565 \nwhich implicitly represent two parallel planes of gold electrode.  566 \n 567 \nA potential is introduce d to represent an applied electric field due to the voltage 568 \ndifference across the two electrodes. The functional form is given in Equation 2: 569 \n 570 \n 𝑈2 = ∑ −𝑞𝑖𝐸𝑧𝑖\n𝑁𝑎𝑡𝑜𝑚𝑠\n𝑖=1\n= ∑ −𝑞𝑖 ( 𝑉\n𝑧0 + 2𝑙𝑆−𝐴𝑢\n)𝑧𝑖\n𝑁𝑎𝑡𝑜𝑚𝑠\n𝑖=1\n (2) \n 571 \nwhere Natoms is the total number of atoms in each system including solvent, qi is the 572 \ncharge of atom i, zi is the z-coordinate of atom i, z0 is the equilibrium end- to-end 573 \ndistance (displacement along z) for a holding stage, and lS-Au is the length of the sulfur-574 \ngold bond.  575 \n 576 \nWe further introduce a potential to orient each sulfur atom’s lone pairs in either the 577 \npositive or negative z-direction, such that a feasible dative bond may occur between the 578 \nsulfur and a fictitious gold particle. This is a key step in ensuring that any conformation 579 \ngenerated by MD simulations can be placed into a gold-gold junction for subsequent 580 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 14 \nNEGF-DFT calcu lations. Because electron lone pairs are not explicitly represented in 581 \natomistic MD simulations, we define surrogate vectors that involve each sulfur’s 582 \nadjacently bonded carbon atoms to act as a proxy for the direction of the electron lone 583 \npairs (Supplementary Figure 17b). We impose a restraint directly on the dot product of 584 \neach surrogate vector with the pulling axis. The functional form of this potential is shown 585 \nin Equation 3: 586 \n 587 \n𝑈3 = ∑ 𝑘3 [|𝑟 ⃗𝑆𝑖 − (\n𝑟 ⃗𝐶𝐺𝑖 − 𝑟 ⃗𝐶𝐸𝑖\n2 )| ∙ 𝑧 ⃗]\n2\n𝑖=1\n= ∑ 𝑘3[|𝑝 ⃗𝑖| ∙ 𝑧 ⃗](−1)𝑖\n2\n𝑖=1\n \n 588 \nwhere 𝑟 ⃗𝑆𝑖 represents the three-dimensional Cartesian coordinates of the sulfur atom of 589 \ninterest, with S 1 and S 2 subscripts indicating the identity of the sulfur atoms in the N-590 \nterminal and C-terminal methionine residues, respectively, 𝑟 ⃗𝐶𝐺𝑖 and 𝑟 ⃗𝐶𝐸𝑖 are Cartesian 591 \ncoordinates of the adjacent carbon atoms covalently bonded to each sulfur of interest, 592 \nand 𝑧 ⃗ is the unit vector in the direction of the z-axis. Vertical lines denote vector 593 \nnormalization. The final term in the equation determines the sign of the potential (and 594 \nthus the direction of the surrogate vector) allowing for one sulfur’s lone pair  to be 595 \noriented in the positive z-direction while the other is oriented oppositely in the negative 596 \nz-direction. The value of k3 is taken as 10 kcal/mol, resulting in a strong potential that 597 \ntightly secures the orientation of sulfur lone pairs towards the implicitly represented gold 598 \nelectrodes (Supplementary Figures 17c,e).   599 \n 600 \nPrincipal component analysis of MD trajectories 601 \nThe resulting MD trajectory data was subjected to dimensionality reduction by means of 602 \nprincipal components analysis (PCA). PCA was performed separately for the tetra- and 603 \npentapeptides simulations. For the tetrapeptides, the Cartesian coordinates of the 604 \npeptide backbone heavy atoms were extracted. The Euclidean distance matrix upper 605 \ntriangle was computed for these 17 shared backbone atoms, resulting in a 136-606 \ndimensional vector representation for each trajectory frame. These vector 607 \nrepresentations, concatenated across all sequences and holding stages and each 608 \ninteratomic distance, were standardized with Z-score normalization. Finally, the first two 609 \nprincipal components were calculated with PCA-whitening using the scikit-learn python 610 \npackage79. PCA of the pentapeptide  trajectories was performed following that of the 611 \ntetrapeptides, with the exception that the shared molecular subgraph of the 612 \npentapeptides was instead composed of 21 backbone heavy atoms, resulting in a 210-613 \ndimensional vector representation for each MD trajectory frame. All other steps were 614 \nperformed identically.  615 \n 616 \nNEGF-DFT calculations 617 \nNEGF-DFT calculations are performed with a DFT based non- equilibrium Green’s 618 \nfunction (NEGF) approach using the TranSiesta and Tbtrans package 58,80,81. The 619 \nelectrodes contain 8 layers of 16 gold atoms along with a pyramid of 10 Au atoms . 620 \nSulfur atoms in the oligopeptide were made to interact with the gold atoms using a 621 \ntrimer binding motif, as described in literature 30. Geometry relaxation of the sequences 622 \nwere performed using generalized gradient approximation-Perdew-Burke -Ernzerhof 623 \n(GGA-PBE) functional82 using the TranSiesta package 58.  SZP basis sets were used for 624 \n(3) \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 15 \nall the gold atoms. DZP basis sets were used for carbon, hydrogen, oxygen, sulfur, and 625 \nnitrogen. Electrode calculations were carried out with a 4 × 4 × 50 k-mesh. The 626 \ngeometry relaxation was carried out using a 4 × 4 × 1 k-mesh, which was performed till 627 \nall the forces were < 0.05 eV/Å. After the junction was relaxed, the transport calculations 628 \nwere carried out using the TranSiesta package 80,81 with the same functionals, basis 629 \nsets, pseudopotential, and k-mesh as the geometry relaxation. Tbtrans 81 was used to 630 \ncarry out the NEGF calculations and to obtain electron transmission as a function of 631 \nenergy (relative to the fermi energy level). NEGF calculations were carried out from - 5 632 \neV to 5 eV with 0.05 eV energy increments . The transmission plots are shifted with 633 \nrespect to the Fermi energy values of each peptide. The difference between charged 634 \nand uncharged species for the MAAM turn configuration (this is a trial sequence, and 635 \nnot the same sequence obtained from the MD simulations using PCA) has been 636 \nreported in Supplementary Fig. 27 . There are similar qualitative agreements between 637 \ncharged (zwitterionic) and uncharged species. 638 \n 639 \nPDOS calculations were carried out for the peptides in the molecular junctions from - 5 640 \neV to 5 eV using Siesta 58. The PDOS calculations were carried out using two Au 641 \npyramids and two Au layers, repeated periodically. The PDOS calculations are carried 642 \nout and plotted over a suitable energy range such that the Fermi energy level of eac h 643 \npeptide falls within the interval. For the PDOS calculations, the plane-wave orbitals are 644 \nprojected into atomic orbitals, and the resulting projection coefficients and atomic orbital 645 \noverlaps that correspond to a given value of the energy in the plot are multiplied 646 \ntogether and summed over for each atom of interest. Orbital visualizations were carried 647 \nout for the molecule with one gold atom on each side using Siesta 58. The orbitals were 648 \nvisualized using Vesta59 to plot HOMO, HOMO-1, LUMO, and LUMO+1 energy levels.  649 \n 650 \nCorresponding author 651 \nFurther information and requests for resources should be directed to and will be fulfilled 652 \nby the lead contact Charles M. Schroeder (cms@illinois.edu).  653 \n 654 \nData Availability  655 \nSolvent-stripped molecular dynamics trajectories are available at:  656 \nhttps://doi.org/10.5281/zenodo.7843691 657 \nAll other data are available from the corresponding author upon request. 658 \n 659 \nCode Availability 660 \nSTM-BJ data were acquired using a custom instrument controlled by custom software 661 \n(Igor Pro, Wavemetrics). Codes for MD simulations and analysis are available at:  662 \ngithub.com/moeenmeigooni/peptide-conductance 663 \n 664 \n 665 \n 666 \n 667 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 16 \n 668 \n 669 \n 670 \n 671 \n 672 \n 673 \nReferences  674 \n1. Stillman, M. Biological Inorganic Chemistry. Structure and Reactivity. Edited by 675 \nIvano Bertini, Harry B. Gray, Edward I. Stiefel and Joan S. Valentine. Angew. Chem. 676 \nIntl. Ed. 46, 8741–8742 (2007). 677 \n2. Winkler, J. R. & Gray, H. B. Long-range electron tunneling. J. Am. Chem. Soc, 136, 678 \n2930–2939 (2014). 679 \n3. Amdursky, N. et al. Electronic Transport via Proteins. Adv. Mat. 26, 7142–7161 680 \n(2014). 681 \n4. Gray, H. B. & Winkler, J. R. Electron tunneling through proteins. Qtly. Rev. of 682 \nBiophysics 36, 341–372 (2003). 683 \n5. Fereiro, J. A. et al. Tunneling explains efficient electron transport via protein 684 \njunctions. Proc. Natl Acad. Sci. USA 115, 4577–4583 (2018). 685 \n6. Nocera, D. G., Winkler, J. R., Yocom, K. M., Bordignon, E. & Gray, H. B. Kinetics of 686 \nIntramolecular Electron Transfer from Ru11 to FeIH in Ruthenium-Modified 687 \nCytochrome c. J. Am. Chem. Soc 106, 5145-5150 (1984). 688 \n7. Shipps, C. et al. Intrinsic electronic conductivity of individual atomically resolved 689 \namyloid crystals reveals micrometer-long hole hopping via tyrosines. Proc. Natl 690 \nAcad. Sci. USA 118, e2014139118 (2021) 691 \n8. Záliš, Stanislav, et al. Photoinduced hole hopping through tryptophans in 692 \nproteins. Proc Natl Acad Sci USA 118, e2024627118 (2021) 693 \n9. Wang, F. et al. Structure of Microbial Nanowires Reveals Stacked Hemes that 694 \nTransport Electrons over Micrometers. Cell 177, 361-369 (2019). 695 \n10. Ru, X., Zhang, P. & Beratan, D. N. Assessing Possible Mechanisms of Micrometer-696 \nScale Electron Transfer in Heme-Free Geobacter sulfurreducens Pili. J.  Phys. 697 \nChem. B 123, 5035–5047 (2019). 698 \n11. Dahl, P. J. et al. A 300-fold conductivity increase in microbial cytochrome nanowires 699 \ndue to temperature-induced restructuring of hydrogen bonding networks. Sci. Adv. 8, 700 \neabm7193 (2022). 701 \n12.  Williamson, H. R., Dow, B. A. & Davidson, V. L. Mechanisms for control of biological 702 \nelectron transfer reactions. Bioorg. chem. 57, 213–221 (2014). 703 \n13.  Hines, T. et al. Transition from tunneling to hopping in single molecular junctions by 704 \nmeasuring length and temperature dependence. J. Am. Chem. Soc. 132, 11658–705 \n11664 (2010). 706 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 17 \n14. Zhang, Y. et al.  Biological charge transfer via flickering resonance. PNAS 111, 707 \n10049-10054 (2014).  708 \n15. Li, Songsong, et al. Transition between nonresonant and resonant charge transport 709 \nin molecular junctions. Nano letters 21, 8340-8347 (2021).  710 \n16. Cordes, M. & Giese, B. Electron transfer in peptides and proteins. Chem. Soc. Rev. 711 \n38, 892–901 (2009). 712 \n17. Juhaniewicz, J., Pawlowski, J. & Sek, S. Electron Transport Mediated by Peptides 713 \nImmobilized on Surfaces. Israel J. Chem. 55, 645–660 (2015). 714 \n18. 16. Scullion, L. et al. Large conductance changes in peptide single molecule 715 \njunctions controlled by pH. J. Phys. Chem. C 115, 8361–8368 (2011). 716 \n19. Baghbanzadeh, M. et al. Charge Tunneling along Short Oligoglycine Chains. Angew. 717 \nChem. 127, 14956–14960 (2015). 718 \n20. Juhaniewicz, J. & Sek, S. Peptide molecular junctions: Distance dependent electron 719 \ntransmission through oligoprolines. Bioelectrochemistry 87, 21–27 (2012). 720 \n21. Guo, Cunlan, et al. Tuning electronic transport via hepta-alanine peptides junction by 721 \ntryptophan doping. PNAS 113, 10785-10790 (2016). 722 \n22.  Xiao, Bingqian Xu, and Tao. Conductance titration of single-peptide 723 \nmolecules. JACS 126 5370-5371 (2004).  724 \n23. Malak, R. A., Gao, Z., Wishart, J. F. & Isied, S. S. Long-range electron transfer 725 \nacross peptide bridges: The transition from electron superexchange to hopping. J. 726 \nAm. Chem. Soc. 126, 13888–13889 (2004). 727 \n24. Marcus, R. A., Sutin, N. & Amos, A. Electron transfers in chemistry and biology. 728 \nBiochimica et Biophysica Acta 811, 265-322 (1985). 729 \n25. Beratan, David N., J. N. Betts, and J. N. Onuchic. Protein electron transfer rates set 730 \nby the bridging secondary and tertiary structure. Science 252, 1285-1288 (1991). 731 \n26. Amdursky, N. Electron Transfer across Helical Peptides. ChemPlusChem 80, 1075–732 \n1095 (2015). 733 \n27. Sepunaru, L. et al. Electronic transport via homopeptides: The role of side chains 734 \nand secondary structure. J. Am. Chem. Soc. 137, 9617–9626 (2015). 735 \n28.  Mandal, H. S. & Kraatz, H. B. Electron transfer mechanism in helical peptides. J. 736 \nPhys. Chem. Lett. 3, 709–713 (2012). 737 \n29. Horsley, J. R., Yu, J., Moore, K. E., Shapter, J. G. & Abell, A. D. Unraveling the 738 \ninterplay of backbone rigidity and electron rich side-chains on electron transfer in 739 \npeptides: The realization of tunable molecular wires. J. Am. Chem. Soc. 136, 740 \n12479–12488 (2014). 741 \n30. Brisendine, J. M. et al. Probing Charge Transport through Peptide Bonds. J. Phys. 742 \nChem. Lett. 9, 763–767 (2018). 743 \n31. Stefani, D. et al. Conformation-dependent charge transport through short peptides. 744 \nNanoscale 13, 3002-3009 (2021). 745 \n32. Lin, Jianping, and David N. Beratan. Tunneling while pulling: the dependence of 746 \ntunneling current on end-to-end distance in a flexible molecule.  J. Phys. Chem 747 \nA 108, 5655-5661 (2004). 748 \n33.  Schneebeli, Severin T., et al. Single-molecule conductance through multiple π− π-749 \nstacked benzene rings determined with direct electrode-to-benzene ring 750 \nconnections. JACS 133, 2136-2139 (2011).  751 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 18 \n34. Zhang, Bintian, et al. Role of contacts in long-range protein conductance. PNAS 752 \n116, 5886-5891 (2019).  753 \n35. Kretchmer, Joshua S., et al. Fluctuating hydrogen-bond networks govern anomalous 754 \nelectron transfer kinetics in a blue copper protein. PNAS 115, 6129-6134 (2018).  755 \n36. Ciudad, S. et al. Aβ(1-42) tetramer and octamer structures reveal edge conductivity 756 \npores as a mechanism for membrane damage. Nat. Commun. 11, 3014-3028 757 \n(2020). 758 \n37. Biswas, M., Lickert, B. & Stock, G. Metadynamics Enhanced Markov Modeling of 759 \nProtein Dynamics. J. Phys. Chem. B 122, 5508–5514 (2018). 760 \n38. Brooks, B. R. et al. CHARMM: The biomolecular simulation program. J. Comput. 761 \nChem. 30, 1545–1614 (2009). 762 \n39. Huang, J. et al. CHARMM36m: An improved force field for folded and intrinsically 763 \ndisordered proteins. Nat. Methods 14, 71–73 (2016). 764 \n40. Wang, H. & Leng, Y. Gold/Benzenedithiolate/Gold Molecular Junction: A Driven 765 \nDynamics Simulation on Structural Evolution and Breaking Force under Pulling. J. 766 \nPhys. Chem. C 119, 15216–15223 (2015). 767 \n41. Mejía, L., Renaud, N. & Franco, I. Signatures of Conformational Dynamics and 768 \nElectrode-Molecule Interactions in the Conductance Profile during Pulling of Single-769 \nMolecule Junctions. J. Phys. Chem. Lett. 9, 745–750 (2018). 770 \n42. Strange, M., Lopez-Acevedo, O. & Häkkinen, H. Oligomeric gold-thiolate units define 771 \nthe properties of the molecular junction between gold and benzene dithiols. J. Phys. 772 \nChem. Lett. 1, 1528–1532 (2010). 773 \n43. Li, Z. & Franco, I. Molecular Electronics: Toward the Atomistic Modeling of 774 \nConductance Histograms. J. Phys. Chem. C 123, 9693–9701 (2019). 775 \n44. Batra, A. et al. Tuning rectification in single-molecular diodes. Nano Lett. 13, 6233–776 \n6237 (2013). 777 \n45. Kumar, P., Paterson, N. G., Clayden, J., & Woolfson, D. N., De novo design of 778 \ndiscrete, stable 310-helix peptide assemblies. Nature, 607(7918), 387-392 (2022).  779 \n46. Brown, R. A., Marcelli, T., De Poli, M., Solà, J., & Clayden, J. (2012). Induction of 780 \nunexpected left‐handed helicity by an N‐terminal L‐amino acid in an otherwise 781 \nachiral peptide chain. Angewandte Chemie, 124(6), 1424-1428 (2012).  782 \n47. Wei, Y., Thyparambil, A. A., & Latour, R. A. Protein helical structure determination 783 \nusing CD spectroscopy for solutions with strong background absorbance from 190 to 784 \n230 nm. (BBA)-Proteins and Proteomics, 1844(12), 2331-2337 (2014).  785 \n48. Li, S. et al. Charge Transport and Quantum Interference Effects in Oxazole-786 \nTerminated Conjugated Oligomers. J. Am. Chem. Soc. 141, 16079–16084 (2019). 787 \n49. Li, B. et al. Intrachain Charge Transport through Conjugated Donor-Acceptor 788 \nOligomers. ACS Appl. Electron. Mater. 1, 7–12 (2019). 789 \n50. Hybertsen, M. S. et al. Amine-linked single-molecule circuits: Systematic trends 790 \nacross molecular families. J. of Phys. Condens. Matter. 20, 374115 (2008). 791 \n51. Bamberger, N. D., Ivie, J. A., Parida, K. N., McGrath, D. v. & Monti, O. L. A. 792 \nUnsupervised Segmentation-Based Machine Learning as an Advanced Analysis 793 \nTool for Single Molecule Break Junction Data. J. Phys. Chem. C 124, 18302–18315 794 \n(2020). 795 \n52. Lin, L. et al. Spectral clustering to analyze the hidden events in single-molecule 796 \nbreak junctions. J. Phys. Chem. C 125, 3623–3630 (2021). 797 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 19 \n53. Liu, B., Murayama, S., Komoto, Y., Tsutsui, M. & Taniguchi, M. Dissecting Time-798 \nEvolved Conductance Behavior of Single Molecule Junctions by Nonparametric 799 \nMachine Learning. J. Phys. Chem. Lett. 11, 6567–6572 (2020). 800 \n54. Cabosart, D. et al. A reference-free clustering method for the analysis of molecular 801 \nbreak-junction measurements. Appl. Phys. Lett. 114, 143102 (2019). 802 \n55. Rousseeuw, P. J. Silhouettes: a graphical aid to the interpretation and validation of 803 \ncluster analysis. J. Compu. and App. Math. 20, 53-65 (1987). 804 \n56. Hollingsworth, Scott A. and Karplus, P. Andrew. \"A fresh look at the Ramachandran        805 \nplot and the occurrence of standard structures in proteins\" Biomolecular Concepts 1, 806 \n271-283 (2010)   807 \n57. Kabsch, W. & Sander, C. Dictionary of protein secondary structure: Pattern 808 \nrecognition of hydrogen‐bonded and geometrical features. Biopolymers 22, 2577–809 \n2637 (1983). 810 \n58. Soler, J. M. et al. The SIESTA method for ab initio order-N materials simulation. J. 811 \nPhys.: Condens. Matter. 14, 2745-2781 (2002). 812 \n59. Momma, K. & Izumi, F. VESTA 3 for three-dimensional visualization of crystal, 813 \nvolumetric and morphology data. J Appl Crystallogr 44, 1272–1276 (2011). 814 \n60. Page, Christopher C., et al. Natural engineering principles of electron tunnelling in 815 \nbiological oxidation–reduction. Nature 402, 47-52 (1999).  816 \n61. Beratan, D. N., Onuchic, J. N. & Hopfield, J. J. Electron tunneling through covalent 817 \nand noncovalent pathways in proteins. J Chem. Phys. 86, 4488–4498 (1987). 818 \n62. Wuttke, Deborah S., et al. Electron-tunneling pathways in cytochrome 819 \nc. Science 256, 1007-1009 (1992).  820 \n63. Onuchic, J. N. & Beratan, D. N. A predictive theoretical model for electron tunneling 821 \npathways in proteins. J Chem. Phys. 92, 722–733 (1990). 822 \n64. Beratan, D. N., Onuchic, J. N., Winkler, J. R. & Gray, H. B. Electron-Tunneling 823 \nPathways in Proteins. Science 258, 1740–1741 (1992). 824 \n65. Betts, J. N., Beratan, D. N., & Onuchic, J. N. Mapping electron tunneling pathways: 825 \nan algorithm that finds the\" minimum length\"/maximum coupling pathway between 826 \nelectron donors and acceptors in proteins. JACS, 114(11), 4043-4046 (1992).  827 \n66. Inkpen, M. S. et al. Non-chemisorbed gold–sulfur binding prevails in self-assembled 828 \nmonolayers. Nat. Chem. 11, 351–358 (2019). 829 \n67. Venkataraman, Latha, et al. Single-molecule circuits with well-defined molecular 830 \nconductance. Nano letters 6, 458-462 (2006).  831 \n68. Hostert, J. D. et al. Self-Assembly and Rearrangement of a Polyproline II Helix 832 \nPeptide on Gold. Langmuir 37, 6115–6122 (2021). 833 \n69. Creasey, R. C. G. et al. Biomimetic Peptide Nanowires Designed for Conductivity. 834 \nACS Omega 4, 1748–1756 (2019). 835 \n70. Nagahara, L. A., Thundat, T. & Lindsay, S. M. Preparation and characterization of 836 \nSTM tips for electrochemical studies. Rev. Sci. Instr. 60, 3128–3130 (1989). 837 \n71. Tien, M. Z., Sydykova, D. K., Meyer, A. G., & Wilke, C. O. PeptideBuilder: A simple 838 \nPython library to generate model peptides. PeerJ 1, e80 (2013).  839 \n72.  Humphrey, W., Dalke, A., & Schulten, K. VMD: visual molecular dynamics. J. mol. 840 \ngraphics 14, 33-38 (1996).  841 \n73. Eastman, P. et al. OpenMM 7: Rapid development of high performance algorithms 842 \nfor molecular dynamics. PLoS Comput. Biol. 13, e1005659 (2017). 843 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 20 \n74. Zhang, Z., Liu, X., Yan, K., Tuckerman, M. E. & Liu, J. Unified Efficient Thermostat 844 \nScheme for the Canonical Ensemble with Holonomic or Isokinetic Constraints via 845 \nMolecular Dynamics. J. Phys. Chem. A 123, 6056–6079 (2019). 846 \n75. Darden, T., York, D. & Pedersen, L. Particle mesh Ewald: An N·log(N) method for 847 \nEwald sums in large systems. J. Chem. Phys. 98, 10089–10092 (1993). 848 \n76.  Scherer, M. K. et al. PyEMMA 2: A Software Package for Estimation, Validation, 849 \nand Analysis of Markov Models. J. Chem. Theory Comput. 11, 5525–5542 (2015). 850 \n77. David shaw, by E. et al. Anton, a Special-Purpose Machine for Molecular Dynamics 851 \nSimulation. Commun. of the ACM 51, 91-97 (2008). 852 \n78. Shirts, M. & Pande, V. S. Screen savers of the world unite. Science 290 1903–1904 853 \n(2000). 854 \n79. Pedregosa, Fabian, et al. \"Scikit-learn: Machine learning in Python. The J.  MLR 12, 855 \n2825-2830 (2011)  856 \n80. Brandbyge, M., Mozos, J. L., Ordejón, P., Taylor, J. & Stokbro, K. Density-functional 857 \nmethod for nonequilibrium electron transport. Phys. Rev. B 65, 1654011–16540117 858 \n(2002). 859 \n81. Papior, N., Lorente, N., Frederiksen, T., García, A. & Brandbyge, M. Improvements 860 \non non-equilibrium and transport Green function techniques: The next-generation 861 \nTRANSIESTA. Comput. Phys. Commun. 212, 8–24 (2017). 862 \n82. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized Gradient Approximation Made 863 \nSimple. Phys. Rev. lett. 77, 3865-3869 (1996).  864 \n 865 \n 866 \n  867 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 21 \nFigure Captions 868 \nFigure 1: Schematic of experimental setup and chemical structures of peptides studied 869 \nin this work. Structures of tetra- and pentapeptide with (a) nonpolar aliphatic, (b) 870 \naromatic, or (c) sterically constrained R-groups. (d) Schematic of a single-molecule 871 \njunction containing peptide with sequence Met- Ala-Ala-Met (MAAM) using 872 \nconformations from MD simulations. (e) Characteristic single-molecule trace for a 873 \npeptide with sequence MAAM revealing two distinct conductance populations.  874 \nFigure 2: Scanning tunneling microsco pe-break junction (STM-BJ) measurements of 875 \noligopeptides at 250 mV applied bias. (a), (b) Characteristic single-molecule traces for 876 \ntetra- and pentapeptides. (c), (d)  1D conductance histograms for the tetra- and 877 \npentapeptides. (e), (f) 2D conductance histograms for MGGM and MGGGM. 878 \nFigure 3: Molecular dynamics (MD) simulations methodology and results. (a) Inter-879 \nanchor displacement potentials  at 6  Å, 9  Å, and 12 Å holding stages defined in Eq. 1 . 880 \n(b) Schematic of applied electric field defined in Eq. 2. (c) Sulfur-orienting potential 881 \ndefined in Eq. 3. (d), (e) Violin plots showing backbone H-bonding distance distribution 882 \nfor tetra- and pentapeptides indicating elimination of intramolecular H-bonds at larger 883 \ndisplacements (12 Å). Triangles indicate peaks in the molecular extension distribution at 884 \nwhich peptide conformers are  selected for NEGF-DFT calculations . (f), (g) Snapshots 885 \nfor tetra- and pentapeptide conformations at small (blue, 6 Å) and large (green, 12 Å ) 886 \ndisplacements.  887 \nFigure 4: Principal component analysis (PCA) results showing the effect of end- to-end 888 \nstretching on peptide conformation and molecular descriptors. (a) Structure of MAAM 889 \nturn conformation from MD simulations shown in a single-molecule junction. (b), (e)  890 \nPrincipal component projections for M AAM turn conformational landscapes denoted 891 \nwith respect to energy and H-bonding distance at a 6 Å holding stage . (d)  Structure of 892 \nMAAM extended conformation from MD simulations shown in a single-molecule 893 \njunction. (c), (f)  Principal component projections for M AAM extended conformational 894 \nlandscapes colored with respect to energy and H-bonding distance a 12 Å holding 895 \nstage. Regions of conformational space corresponding to low  i → i+3 backbone H-bond 896 \ndistances are shown to deplete with increasing inter-anchor displacements as denoted 897 \nby the black dotted circle. Blue and green crosses correspond to MAAM turn and 898 \nextended conformations, respectively.  899 \nFigure 5: Non-equilibrium Green’s function -density functional theory (NEGF-DFT) 900 \ncalculations for electron transport. (a), (b)  Schematic of molecular junctions showing 901 \ngold metal electrodes and MGGM turn and extended conformations for NEGF-DFT 902 \ncalculations. (c), (d) Transmission probability as a function of energy (relative to the 903 \nFermi energy level) for MGGM and MGGGM, showing drastic differences in 904 \ntransmission probability at E - E F = 0 for the turn (blue) and the extended (green) 905 \nconfigurations. (e), (f) Zero bias conductance for tetra- and pentapeptides indicating 906 \nlarge differences in transmission probabilities between the two conformational states.   907 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n 22 \nAcknowledgments 908 \nThis work was supported by the U.S. Department of Energy, Office of Science, Basic 909 \nEnergy Sciences under Award No. DE-SC0022035 for M. Meigooni, H.Y., J.L., E.T. , 910 \nX.L, and C.M.S. and the National Science Foundation under Award 2227399 for R.S. 911 \nand C.M.S. 912 \nAuthor contributions 913 \nR.S., M. Meigooni, and C.M.S. conceived this study. R.S. performed STM-BJ 914 \nexperiments and NEGF-DFT calculations. M. Meigooni performed MD simulations and 915 \nPCA analysis. H.Y. assisted with control experiments and J.L. assisted with GMM 916 \nmodeling. X.L assisted with the CD meas urements. M. Mosquera and N.E.J. assisted 917 \nwith the NEGF-DFT calculations. E.T. and C.M.S. supervised the research. The 918 \nmanuscript was written by R.S., M. Meigooni, E.T. and C.M.S. with contributions from all 919 \nauthors.  920 \nCompeting Interests  921 \nThe authors declare no competing interests.  922 \nAdditional Information 923 \nSupplementary information contains supplementary figures, supplementary tables, and 924 \nsupplementary text. 925 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n \na \nb \nc \nd \ne \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\na\nb \nc \nd \ne \nf \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n \na \n c\nb\nd \n e \nf \n g \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n \na \n b \n c \n d \ne \n f \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint \n\n \na \nb\n \nc \nd\n \ne\n \nf\n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}