A Tutorial on Automated Mass Spectral Analysis Using NucleoSpec for Compositional Assignment of Nucleic Acid-Silver Complexes and Nanoclusters

Abstract

Nucleic acid–metal complexes, such as DNA-stabilized silver nanoclusters (DNA-Ag _N ) and Ag­(I)-mediated DNA duplexes, show promise for applications ranging from biomedical imaging to bioelectronics and sensing. Electrospray ionization mass spectrometry (ESI-MS) is a powerful tool for precisely resolving their compositions, but the data analysis procedure can be challenging and time-consuming. No general-purpose software can automatically assign compositions of nucleic acid–silver complexes, most notably for DNA-Ag _N , whose masses depend on the oxidation state of the silver core. We address these challenges by developing an automated data analysis workflow for ESI-MS of nucleic acid–silver complexes. This tutorial presents the analysis workflow and introduces NucleoSpec, an open-source, web-based platform that consolidates charge state determination, constraint-based composition enumeration, isotope-pattern matching, and adduct/conjugate handling into a single browser-accessible interface. The platform is benchmarked on published molecular formulas determined by manual analysis for DNA-Ag _N , xeno-nucleic acid stabilized Ag _N (XNA-Ag _N ), nucleic acid–Ag­(I) complexes, and chemically modified DNA-Ag _N , reproducing assignments to single-proton precision (median centroid offset of 0.03 m/z) for spectra of adequate signal-to-noise. Low-signal peaks for which adjacent valence electron counts become difficult to distinguish are flagged by NucleoSpec for user inspection. By lowering the barrier to high-resolution compositional analysis and supporting integration into high-throughput pipelines, NucleoSpec addresses a key experimental bottleneck in informatics-guided design of nucleic acid–metal complexes.

1. Introduction

The affinity of nucleobases for silver is a useful tool for expanding sequence-programmed DNA nanotechnologies. For example, Ag­(I) can mediate noncanonical duplexes with significantly increased stabilities and can expand the sequence rules that govern base pairing in natural DNA. − DNA can also stabilize silver nanoclusters (DNA-Ag _N ) with atomically precise sizes and photoluminescence properties that depend sensitively on DNA sequence. − DNA-Ag _N exhibit a diversity of emission wavelengths spanning the visible to near-infrared range and have been applied in fluorescence imaging and biosensing. − To enable the development of applications that utilize such DNA-metal complexes, it is critical to understand the compositions, structures, and properties of these complexes.

Mass spectrometry of samples purified by chromatography has been essential for understanding DNA-Ag _N structure–property relationships and for shedding light on nucleic acid-Ag­(I) complexes. In particular, high-resolution, negative-ion mode electrospray ionization mass spectrometry (ESI-MS) is the best-suited tool for determining DNA-Ag _N composition. ,, This approach can be used to resolve nucleic acid complexes to single-proton precision, enabling determination of both nanocluster composition and charge by fitting the characteristic isotopic distribution of the resolved product. ,,− When combined with recently reported crystal structures, , this information about DNA-Ag _N molecular formula has enabled the first realistic theoretical calculations of DNA-Ag _N electronic structure , and uncovered halide ligand chemistry for stabilizing these emitters. ESI-MS is also an extremely valuable tool for probing the compositions of DNA-Ag­(I) complexes. ,,,, More broadly, ESI-MS has emerged as a powerful platform for elucidating growth mechanisms and capturing transient intermediates in atomically precise metal nanocluster synthesis, , further motivating the development of automated analysis tools that can keep pace with the expanding scale and complexity of such studies.

It remains a time-consuming manual process to accurately extract the compositions of nucleic acid–silver complexes from ESI-MS data. The process can require the use of multiple software tools to validate the correct composition assignment. Also, nucleic acid-silver complexes often fragment under ESI-MS, increasing the numbers of mass spectral peaks that must be analyzed to correctly assign composition of the product of interest. , Moreover, the complete assignment of DNA-Ag _N composition requires more than just determining the number of DNA strands, silver atoms, and other adducts: the oxidation state of the inorganic nanocluster core must also be determined from the proton-resolved mass of the complex. Because charge state modifies the partial protonation of the nucleic acid’s phosphate backbone that is needed to reach a given charge state, this requires the analyst to enforce a domain-specific charge-balance constraint that no general-purpose deconvolution tool is able to enforce. Determination of nanocluster oxidation state is crucial because it enables one to measure the effective valence electron count of the inorganic core, which is a parameter that is strongly correlated to DNA-Ag _N optical properties ,, and is required for accurate modeling of electronic structure. ,

General-purpose mass spectrometry tools such as mMass and OpenMS, which provide deconvolution and isotope-matching capabilities, do not natively incorporate domain-specific constraints that are relevant to nucleic acid–silver nanocluster analysis, such as assigning silver nanocluster oxidation state. General-purpose tools also do not automatically support cross-charge state validation for nanoclusters and are not designed to automatically detect adducts or other chemical modifications in DNA-silver complexes. As high-throughput screening and machine learning expand the scale of DNA-Ag _N discovery, − compositional analysis of nucleic acid-silver complexes by ESI-MS remains a key bottleneck. To date, no automated tool has been developed for compositional analysis of nucleic acid–silver complexes from ESI-MS data, and all published assignments have relied on manual workflows.

This tutorial addresses this bottleneck. Here, we introduce NucleoSpec, an open-source platform for automated compositional analysis of both nucleic acid–stabilized silver nanoclusters and nucleic acid-Ag­(I) complexes from ESI-MS data. The platform integrates isotope-pattern matching, charge state determination, and constraint-based validation within a browser-accessible interface. The analytical framework is validated for DNA-Ag _N , xeno nucleic acid–stabilized silver nanoclusters (XNA-Ag _N ), Ag­(I)-mediated DNA duplexes, , and chemically conjugated DNA-Ag _N for click chemistry. , The framework distinguishes covalent conjugates from noncovalent adducts by incorporating conjugate elemental composition directly into the nucleic acid mass model while treating adducts as charge-bearing mass offsets. This tool significantly decreases the time required to determine the composition of nucleic acid-silver complexes to single-proton resolution, and the tool holds potential for integration into automated workflows for high-throughput experiments. We have implemented NucleoSpec as a web-based platform, and users can also run the open-source tool locally. These features of NucleoSpec lower the barrier for high-resolution mass spectral analysis for those entering the rapidly growing field of DNA-noble metal complexes, ensuring high-quality analytical chemical analysis in a single platform.

2. Background: ESI-MS of Nucleic Acid–Silver Complexes

We first provide a tutorial introduction to the data analysis methodology and workflow for negative ion mode ESI-MS characterization of DNA-Ag _N and Ag­(I)-mediated DNA duplexes, which have been presented together with detailed experimental methods in prior works. ,,,,, Nucleic acids at neutral pH are anions with deprotonated phosphate backbones, and negative ion mode ESI-MS is well-suited for gentle characterization of noncovalent nucleic acid complexes. Ammonium acetate solution is commonly used as a volatile solvent for gentle desolvation of nucleic acids by ESI, which promotes protonation of the negatively charged phosphate backbone by ammonium ions and thereby reduces the charge of the complex and produces volatile ammonia. ,, This process yields a range of products with partially protonated phosphate backbones, and this means that the same mass species can be observed at multiple charge states z that differ only by the degree of backbone protonation. Thus, for precise determination of complex composition, and especially to determine nanocluster oxidation state with high precision, ,, it is important to validate composition at multiple charge states, which should differ only by integer numbers of protons. The focus of this tutorial is on automation of ESI-MS data analysis; for detailed discussion of the experimental methods for ESI-MS of nucleic acid–silver complexes, the reader is referred to prior works. ,,,,,,

2.1. Mass–Composition Relationships

Composition determination for a silver-DNA complex must take into account the contribution of the net charge of the inorganic nanocluster core in the complex. A complex of n _s bare nucleic acids at charge state z has a mass m equal to the un-ionized (i.e., neutralized) nucleic acids minus the protons that must be removed from the un-ionized nucleic acids to reach charge z:

$$m=M_{\mathrm{N}\mathrm{A}}{n}_s-z{m}_{\mathrm{p}}$$ 1

where M _NA is the mass of the un-ionized nucleic acid (i.e., with all phosphates protonated) and m _p is the mass of a proton. If silver cations are present in the nucleic acid complex, these contribute +1 per Ag­(I) to the charge z, thus changing the number of protons that must be removed to reach z. The mass of a nucleic acid-Ag­(I) complex with a total number of silvers n _Ag at charge state z is then:

$$m=M_{\mathrm{N}\mathrm{A}}{n}_s+n_{\mathrm{A}\mathrm{g}}{M}_{\mathrm{A}\mathrm{g}}-(n_{\mathrm{A}\mathrm{g}}+z)m_{\mathrm{p}}$$ 2

where M _Ag is the mass of silver.

If the nucleic acid-silver complex contains a silver nanocluster, high-resolution ESI-MS can be used to determine the charge Q _cl of the partially oxidized inorganic cluster core. This is possible because only cationic silvers will offset the number of protons that must be added to the complex to reach z, which allows one to distinguish between reduced (neutral) silvers and cationic silvers. The total silver count n _Ag of the nanocluster can be separated into the effective numbers of reduced silvers, N ₀, and cationic silvers, N ₊:

$$N_0+N_{+}=n_{\mathrm{A}\mathrm{g}}$$ 3

If no other charged adducts or ligands are present, the charge of the inorganic core is Q _cl = N ₊, and the resolved product at charge state z has mass given by

$$m=M_{\mathrm{N}\mathrm{A}}{n}_s+n_{\mathrm{A}\mathrm{g}}{M}_{\mathrm{A}\mathrm{g}}-(Q_{\mathrm{c}\mathrm{l}}+z)m_{\mathrm{p}}$$ 4

This allows determination of N ₀, which also represents the effective valence electron count of the nanocluster core. N ₀ is a highly important quantity for ab initio calculations of nanocluster electronic structure. ,

Finally, it is possible to observe the presence of charged adducts, including halidos integrated into specific sites in the nanocluster ligand shell that contribute to the electron count of the nanocluster, as well as sodiums and ammoniums that are commonly observed adducts in ESI-MS of nucleic acids. When a mass M _adduct with charge q _adduct is present, Q _cl = N ₊ + q _adduct and the mass of the complex is

$$m=M_{\mathrm{N}\mathrm{A}}{n}_s+n_{\mathrm{A}\mathrm{g}}{M}_{\mathrm{A}\mathrm{g}}+M_{\mathrm{a}\mathrm{d}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}}-(Q_{\mathrm{c}\mathrm{l}}+z)m_{\mathrm{p}}$$ 5

The nucleic acids may carry covalent modifications such as click-chemistry handles. If a covalent conjugate is neutral and is bound to the strand before silver addition, its elemental mass simply augments M _NA without altering the charge-balance bookkeeping that distinguishes N ₀ from N ₊.

3. Getting Started with NucleoSpec

We next introduce an automated workflow, NucleoSpec, for analyzing ESI-MS data for nucleic acid–silver complexes. First, a simple tutorial for the workflow is presented using a representative example for which the complete mass spectrum is analyzed. Then, the algorithmic details of the workflow are described.

3.1. Interface and User Workflow

NucleoSpec can be used through the hosted instance at https://huggingface.co/spaces/copplab/NucleoSpec without installation, or installed locally from the GitHub repository (https://github.com/copplab/NucleoSpec) via conda env create -f environment.yml. Local installation is recommended for high-throughput analysis or for users handling proprietary spectra. Note that the hosted instance does not store user-uploaded data.

A first-time user can analyze a spectrum in four steps, as illustrated in Figure :

1.

Analysis workflow: automated spectral preprocessing, charge state determination, composition enumeration, and isotope-pattern matching with optional adduct/conjugate modeling, followed by user-directed cross-charge state validation.

(1) Upload a centroided ESI-MS spectrum as a two-column tab- or comma-delimited text file (m/z, intensity). The input spectrum must have sufficient resolving power to separate adjacent isotope peaks at the observed charge states (typically R = m/Δm ≳ 10 000 for z ≤ 7); lower resolution may cause charge-state misassignment.

(2) Select the analysis mode (DNA-Ag _N , XNA-Ag _N , or Ag­(I)-DNA/XNA Complex) and enter the nucleic acid sequence or elemental formula.

(3) Click on any resolved isotope envelope in the interactive spectrum viewer. NucleoSpec automatically determines the charge state, enumerates candidate compositions, and displays ranked results with theoretical isotope overlays.

(4) For species resolved at multiple charge states, click each additional envelope to confirm cross-charge state consistency.

Example spectra, including the DNA-Ag _N sample stabilized by oligonucleotide 5′-ATCCCCTGTC-3′ from Table and Table S5, are provided in the GitHub repository and on the hosted instance to allow immediate replication of the figures in this work. On the hosted instance, users should wait until charge state annotations appear above peaks in the spectrum viewer before clicking, as data processing can take time, depending on internet connectivity and processing power. Users can monitor processing progress in the log panel.

1. Summary of Representative Validation Cases Analyzed Using NucleoSpec .

System	Sequence	m/z	z	NucleoSpec detected composition	N 0	Q cl	Pattern Similarity	ΔX 0	Confidence tier	ref
DNA–Ag N (ssDNA)	ATCCCCTGTC	1924.09	4	$${[{(\mathrm{DNA})}_2{\mathrm{Ag}}_{17}]}^{\mathrm{11+}}$$	6	11	0.999	0.0015	▲
		1539.06	5	$${[{(\mathrm{DNA})}_2{\mathrm{Ag}}_{17}]}^{\mathrm{11+}}$$	6	11	0.999	0.012	▲
		1282.39	6	$${[{(\mathrm{DNA})}_2{\mathrm{Ag}}_{17}]}^{\mathrm{11+}}$$	6	11	0.993	0.015	▲
DNA–Ag N with adduct	CACCTAGCGA	2533.51	3	$${[{(\mathrm{DNA})}_2{\mathrm{Ag}}_{15}]}^{\mathrm{9+}}$$	6	9	0.996	0.039	▲
		1899.64	4	$${[{(\mathrm{DNA})}_2{\mathrm{Ag}}_{15}]}^{\mathrm{9+}}$$	6	9	0.996	0.0034	▲
		1944.34	4	$${[{(\mathrm{DNA})}_2{\mathrm{Ag}}_{16}{\mathrm{Cl}}_2]}^{\mathrm{8+}}$$	6	8	0.996	0.046	▲
		1555.50	5	$${[{(\mathrm{DNA})}_2{\mathrm{Ag}}_{16}{\mathrm{Cl}}_2]}^{8+}$$	6	8	0.999	0.022	▲
(DNA-BCN) n –Ag N	CACCTAGCGA	2138.69	4	$${[{(\mathrm{DNA}‐\mathrm{BCN})}_2{\mathrm{Ag}}_{16}{\mathrm{Cl}}_2]}^{6+}$$	8	6	0.975	0.014	▲
		1710.34	5	$${[{(\mathrm{DNA}‐\mathrm{BCN})}_2{\mathrm{Ag}}_{16}{\mathrm{Cl}}_2]}^{\mathrm{6+}}$$	8	6	0.963	0.082	▲
XNA–Ag N (SNA)	SNA-C10	1864.59	4	$${[{(\mathrm{SNA})}_2{\mathrm{Ag}}_{14}]}^{\mathrm{8+}}$$	6	8	0.840	0.075	▲
		1891.07	4	$${[{(\mathrm{SNA})}_2{\mathrm{Ag}}_{15}]}^{\mathrm{9+}}$$	6	9	0.529	0.029	○
XNA–Ag N (TNA)	TNA-C10	1557.43	4	$${[{(\mathrm{TNA})}_2{\mathrm{Ag}}_8]}^{\mathrm{8+}}$$	0	8	0.918	0.020	▲
		1638.59	4	$${[{(\mathrm{TNA})}_2{\mathrm{Ag}}_{11}]}^{\mathrm{7+}}$$	4	7	0.957	0.035	▲
DNA complex	dC20	2651.00	5	$${[{(\mathrm{dC}20)}_2{\mathrm{Ag}}_{17}]}^{\mathrm{17+}}$$	0	17	0.965	0.014	▲
		2208.98	6	$${[{(\mathrm{dC}20)}_2{\mathrm{Ag}}_{17}]}^{\mathrm{17+}}$$	0	17	0.979	0.010	▲
		1893.43	7	$${[{(\mathrm{dC}20)}_2{\mathrm{Ag}}_{17}]}^{\mathrm{17+}}$$	0	17	0.918	0.096	▲
		2714.74	5	$${[{(\mathrm{dC}20)}_2{\mathrm{Ag}}_{20}]}^{\mathrm{20+}}$$	0	20	0.954	0.065	▲
		2262.43	6	$${[{(\mathrm{dC}20)}_2{\mathrm{Ag}}_{20}]}^{\mathrm{20+}}$$	0	20	0.961	0.22	▲
		1939.37	7	$${[{(\mathrm{dC}20)}_2{\mathrm{Ag}}_{20}]}^{\mathrm{20+}}$$	0	20	0.953	0.033	▲
		2280.08	6	$${[{(\mathrm{dC}20)}_2{\mathrm{Ag}}_{21}]}^{\mathrm{21+}}$$	0	21	0.946	0.024	▲
XNA complex (SNA)	SNA-C10	1488.80	2	[(SNA)]	0	0	0.984	0.012	▲
		2092.70	3	$${[{(\mathrm{SNA})}_2{\mathrm{Ag}}_3]}^{\mathrm{3+}}$$	0	3	0.909	0.10	▲
		2100.00	3	$${[{(\mathrm{SNA})}_2{\mathrm{Ag}}_3\mathrm{Na}]}^{\mathrm{4+}}$$	0	4	0.951	0.12	▲
		2128.00	3	$${[{(\mathrm{SNA})}_2{\mathrm{Ag}}_4]}^{\mathrm{4+}}$$	0	4	0.952	0.15	▲
XNA complex (TNA)	TNA-C10	1791.91	3	[(TNA)2]	0	0	0.996	0.079	▲
		1343.67	4	[(TNA)2]	0	0	0.997	0.10	▲
		1584.20	4	$${[{(\mathrm{TNA})}_2{\mathrm{Ag}}_9]}^{\mathrm{9+}}$$	0	9	0.996	0.0041	▲
		2076.60	3	$${[{(\mathrm{TNA})}_2{\mathrm{Ag}}_8]}^{\mathrm{8+}}$$	0	8	0.999	0.0057	▲

^aRepresentative peaks illustrating diverse composition types and charge states; most detected peaks are omitted for brevity. Best-fit selection uses smallest |ΔX ₀| (centroid offset between experimental and theoretical isotope distributions, in m/z). Pattern similarity (0–1) quantifies envelope-shape agreement. For adduct-containing products, Q _cl = n _Ag – N ₀ + q _adduct.

^bConfidence tiers: (▲) high (pattern $>0.8$ ), (○) moderate (0.5–0.8), (▽) low $(<0.5)$ . All compositions match the corresponding published reference.

^cApparent N ₀ shifted by +2 due to in-source +2H modification of BCN alkyne; corrected N ₀ = 6 (see text).

^dLow pattern similarity due to low signal-to-noise ratio.

3.2. Example: Complete Analysis of a DNA-Ag _N Spectrum

To illustrate the workflow, we next walk through the analysis of a representative DNA-Ag _N mass spectrum reported in Guha et al. (Figure ). The reader can reproduce every step below using the sample file, which is preloaded as an example on the hosted instance of the workflow, and is also provided in the GitHub repository.

• 1. Load the spectrum. On the hosted instance, click the “Load sample”; one can also upload the file from the GitHub sample_data/ directory. Loading the sample autopopulates the analysis mode (“DNA-Ag _N ”) and the sequence (ATCCCCTGTC). For a user-supplied spectrum, these would be entered manually. NucleoSpec then renders the spectrum and automatically computes the monoisotopic mass of the neutral strand.
• 2. Wait for charge state annotations. NucleoSpec performs automated peak detection and labels each resolved envelope with its inferred charge state (labels are visible above the peaks in Figure a).
• 3. Click the dominant envelope at m/z ≈ 1924. Within ∼1 s, the candidate composition panel populates with ranked assignments. The top-ranked composition is (DNA)₂Ag₁₇ with N ₀ = 6 and Q _cl = 11. The isotope-overlay panel displays the experimental envelope together with theoretical envelopes for the top-ranked assignment and its Q _cl ±1 neighbors (Figure b); the best fit is the one whose centroid is closest to the experimental centroid (smallest |ΔX ₀|).
• 4. Confirm by cross-charge state validation. Click the envelopes at m/z ≈ 1539 (z = 5) and m/z ≈ 1282 (z = 6), which correspond to the same product with varying degrees of phosphate backbone protonation. Each returns the same (n _s , n _Ag, Q _cl) = (2, 17, 11) assignment with pattern similarity 0.993–0.999 and |ΔX ₀| < 0.02 m/z (Table , Figure c).

2.

NucleoSpec example analysis for DNA-Ag _N species. (a) Experimental ESI-MS spectrum of a DNA-Ag _N species stabilized by DNA oligomer with sequence 5′-ATCCCCTGTC-3′. NucleoSpec labels peaks by charge state. (b) Experimental and theoretical isotope envelope overlay for the top-ranked composition alongside Q _cl ± 1 alternatives. The best-fit cluster charge is selected as that with the smallest ΔX ₀ (i.e., the closest centroid match). The isotope envelope overlay confirms the assignment visually for the user, showing closest agreement in peak spacing and relative intensity of the assigned “best fit” as compared to adjacent Q _cl values with shifted isotopic envelopes. (c) Cross-charge state validation: the same nanocluster composition is observed at multiple charge states, with consistently low ΔX ₀ across envelopes. These results reproduce the composition reported in the original study and demonstrate the use of chemically motivated constraints and cross-charge state consistency to narrow plausible assignments.

The molecular formula assignment of (DNA)₂Ag₁₇ with N ₀ = 6 and Q _cl = 11 across three charge states (Figure b) reproduces the molecular formula reported in the original study. (Figure c) This illustrates how the combination of constraint-based enumeration and cross-charge state validation narrows the candidate list to a single chemically self-consistent composition. A full peak-by-peak analysis of this spectrum, including the remaining minor species and isotope-overlay comparisons, is provided in Figure S2 and Table S5.

Important tips for the user: a few diagnostic habits are important to make this workflow robust in practice for samples of unknown composition. First, before trusting any assignment, inspect the theoretical overlay: a correct fit should reproduce both the peak spacing (which reflect the charge state) and the overall isotope envelope shape (which is governed by the characteristic ¹⁰⁷Ag/¹⁰⁹Ag isotope abundance). Second, if clicking the same species at different charge states yields different top-ranked Q _cl value, the user should inspect the isotope overlay at each z to determine which assignment is correct. Third, if no candidate appears within the displayed mass tolerance, the most common causes are (i) an incorrect or nondeprotonated sequence/formula input, (ii) a charge-state misassignment, or (iii) an unaccounted-for adduct or conjugate. Checking these three in order resolves the majority of failures.

4. The Algorithmic Pipeline

NucleoSpec consolidates the manual workflow described in Section into three stages: (1) automated spectral preprocessing and charge state determination, (2) constraint-based composition assignment, and (3) optional cross-charge state validation. NucleoSpec has capabilities beyond general-purpose tools, such as mMass and OpenMS, because the workflow encodes three domain-specific constraints: the possibility of mixed neutral/cationic silver population (N ₀ + N ₊ = n _Ag), conjugate-aware elemental formulas that treat covalent modifications as neutral parts of the nucleic acid strand, and cross-charge state consistency checks across independently analyzed peaks.

4.1. Charge-State Determination

After peak detection, the charge state z is inferred from the spacing of adjacent isotopes in each envelope (Δm/z ≈ 1.003/z), with automatic correction for the apparent charge-doubling that polyisotopic silver can produce (Supporting Information S1). For the user: the interface provides a manual charge state override for envelopes that are not assigned confidently. Misassignment can occur for data with low signal-to-noise ratios.

4.2. Constraint-Based Composition Enumeration

Following charge-state determination, candidate compositions are generated by systematic enumeration for a range of numbers of nucleic acid strands (n _s = 1–3) and silver atoms (n _Ag = 0–30); these ranges encompass all nucleic acid–Ag _N stoichiometries reported to date. These upper bounds are defined as global parameters in the source code (MAX_STRANDS and MAX_SILVER in core/analyzer.py) and can be adjusted by a user running a local installation from the GitHub repository to accommodate larger or higher-order complexes, with no algorithmic changes required; these parameters are not exposed in the hosted browser interface. Assignments are constrained by N ₀ + N ₊ = n _Ag and Q _cl = N ₊ + q _adduct. The theoretical mass-to-charge ratio for each candidate is

$$\frac{m}{z}=\frac{M_{\mathrm{N}\mathrm{A}}{n}_s+n_{\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{j}}{M}_{\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{j}}+n_{\mathrm{A}\mathrm{g}}{M}_{\mathrm{A}\mathrm{g}}+M_{\mathrm{a}\mathrm{d}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}}-(Q_{\mathrm{c}\mathrm{l}}+z)m_{\mathrm{p}}}{z}$$ 6

where M _NA is computed from the monoisotopic elemental formula of the neutral (protonated) strand, n _conj M _conj is the mass contribution of n _conj covalent conjugates of mass M _conj, M _adduct is the noncovalent adduct mass, and (Q _cl + z)m _p is the total mass of all protons that must be removed from the un-ionized nucleic acid to produce the observed z ^– ion (both the protons that compensate Q _cl and those required to reach the observed ion charge z). When no conjugate or adduct is present, eq reduces to m/z = (M _NA n _s + n _Ag M _Ag – (N ₊ + z) m _p)/z. A coarse m/z prefilter followed by a tighter ppm tolerance (Supporting Information S2) eliminates the majority of combinations and typically leaves a handful of candidates differing only in N ₊.

When a covalent conjugate is present, e.g., a click-chemistry handle such as bicyclononyne (BCN), M _conj is the neutral mass of the conjugate moiety, and n _conj is the total number of conjugates in the complex. Because the conjugate is covalently attached to the nucleic acid before silver salt is added, its elemental composition is added directly to the nucleic acid formula for both mass and isotope-pattern calculations. NucleoSpec allows n _conj to differ from n _s in order to accommodate for conjugation of only a subset of nucleic acid oligomers in the complex; the current implementation treats the conjugate as electrically neutral, with charged modifications handled through the adduct framework. For the user: if a defined conjugate fails to give a satisfactory centroid match in the fully conjugated case, the algorithm automatically tests partial-conjugation hypotheses before reporting failure (details in Supporting Information S2).

The treatment of conjugate mass in eq differs from that of noncovalent adducts (e.g., commonly observed Na⁺ and ${\mathrm{NH}}_4^{+}$ ) or ionic ligands Cl^– and Br^– that are found within the ligand shell of certain DNA-Ag _N and that carry specific charge and produce charge-dependent mass shifts. Because the relationship Q _cl = N ₊ + q _adduct already accounts for the adduct charge, both the adduct mass and its effect on backbone protonation are captured by the M _adduct and (Q _cl + z)m _p terms in eq .

Recent studies have shown that certain DNA-Ag _N can be detected as dimers of nanoclusters in their ESI-MS spectra. , NucleoSpec is also configured to search for these dimers. For peaks where no monomer candidate fits within tolerance (where “monomer” is defined as a single DNA-Ag _N ), or where all monomer candidates show poor centroid agreement, NucleoSpec automatically tests a dimer hypothesis (two copies of the same nucleic acid–silver cluster, which appear at doubled charge in ESI-MS) and adopts it only when its centroid error is strictly smaller than the best monomer fit. Dimer assignments are flagged in the interface and display the doubled-ion formula alongside the monomer composition values; the underlying derivation and tie-breaking logic are described in Supporting Information S2.

4.3. Peak-Matching Metrics and Confidence Tiers

Theoretical isotope distributions for each surviving candidate are generated using IsoSpecPy and compared with the experimental envelope by two metrics: (1) a pattern similarity score that is the mean of cosine similarity and Pearson correlation between matched peaks, and (2) the centroid offset ΔX ₀ between Gaussian fits to the experimental and calculated isotope envelopes (Supporting Information S4). Because surviving candidates after mass filtering typically differ only by one unit in Q _cl, which is a difference that shifts the centroid by ∼m _p/z while leaving envelope shape nearly unchanged (Table S4, median score gap 0.024), ΔX ₀ is the more discriminating metric for ranking adjacent assignments. For the user, this means: the best-fit composition is selected as the candidate with the smallest |ΔX ₀| among those passing mass-balance validation. Pattern similarity (ranging from 0–1) is reported separately as an independent confidence indicator (Supporting Information S4); rationale and parameter sensitivity are given in Supporting Information S4. The strongest confirmation of the accuracy of an assignment is cross-charge state consistency of the molecular formula: clicking the same species at additional charge states should return the same (n _s , n _Ag, Q _cl), and any disagreement between charge states signals that one or more peaks warrant manual inspection.

To guide interpretation, we classify assignments into confidence tiers based on pattern similarity (Table ): high confidence (▲, pattern similarity $>0.8$ ), where isotope envelopes show close agreement and compositions can be considered highly reliable; moderate confidence (◦, pattern similarity = 0.5 to 0.8), where assignments are plausible but may benefit from cross-charge state validation or inspection by the domain expert; and low confidence (▽, pattern similarity $<0.5$ ), where tentative assignments require expert verification. These thresholds are provided to the user as practical guidelines rather than statistically validated cutoffs, and the thresholds reflect isotope-envelope quality rather than assignment correctness: the correct composition can still be recovered when pattern similarity is low, provided ΔX ₀ remains small.

4.4. Common Pitfalls and How to Recognize Them

The following failure modes are the ones most likely to arise in practice. Each entry lists a diagnostic signature so that the user can distinguish the failure mode from a correct but low-confidence assignment, together with the corrective action.

• 1. Charge-state misassignment (z reported as 2z) For silver-containing species, the isotope envelope at charge state z can exhibit alternating high-low peak intensities. This can be caused when charge state z overlaps with a dimerized product of twice the mass at 2z (DNA-Ag _N have been reported to dimerize under ESI conditions). Apex counting may interpret the intermediate peaks as additional isotope spacings and report 2z instead of the true z (e.g., z = 2 reported as z = 4). Diagnostic: alternating high-low peak intensities across the envelope. Action: NucleoSpec detects and corrects this automatically (Supporting Information S1); if the corrected z still appears inconsistent with the same species at other charge states, use the manual charge state override.
• 2. Sequence or formula input not in the neutral (un-ionized) form. The mass model in eqs – requires the fully protonated nucleic acid mass; supplying a deprotonated formula shifts M _NA and produces systematic ΔX ₀ offsets. Diagnostic: no candidate appears within tolerance, or all candidates carry a ΔX ₀ offset that scales with 1/z. Action: re-enter the strand’s mass in neutral form.
• 3. Unaccounted-for adduct or conjugate. Salt adducts (Na⁺, ${\mathrm{NH}}_4^{+}$ ) are common in negative-mode ESI of nucleic acid–silver complexes, and halido ligands (Cl^–, Br^–) can be found on DNA-Ag _N . Covalent modifications (e.g., click-chemistry handles) must be declared as conjugates rather than left implicit, as NucleoSpec will not automatically search masses containing covalent modifications. Diagnostic: the apparent n _Ag is off by an integer, and the residual mass matches a tabulated adduct (Supporting Information S4, Table S2). Action: enable the relevant adduct in the screening library, or declare the conjugate via the molecular editor.
• 4. In-source chemical modification that shifts apparent N ₀. Some functional groups (notably the BCN alkyne, which gains +2H during ESI,) can modify the conjugate mass and shift the apparent N ₀ by a small integer because NucleoSpec does not account for protonation of covalent conjugates. Diagnostic: the apparent N ₀ differs from the unconjugated analogue by an integer that matches the conjugate modification mass. Action: the user should correct N ₀ by the known modification offset; the cluster core composition should then be the corrected value.
• 5. Inflated ΔX₀ forbare strands or low-signal-to-noise envelopes. For products with few or no silvers, isotope envelopes will lack the Gaussian-like ¹⁰⁷Ag/¹⁰⁹Ag fine structure, and the Gaussian-fit centroid is poorly constrained, producing ΔX ₀ values that do not reflect mass-assignment error. This can also occur for noisy data. Diagnostic: anomalously large ΔX ₀ at one charge state but not others for peaks corresponding to the same composition at varying z. Action: apply cross-charge state validation, i.e., compare the assigned compositions at multiple z, rather than treating ΔX ₀ at a single z as decisive of product composition. In such cases, users should not treat the top-ranked assignment at a single z as decisive; instead, the user should consult the displayed pattern similarity and cross-charge state consistency to identify the correct composition (as illustrated by the silver-free cases in Table ). It should be noted that NucleoSpec is designed for use with silver-nucleic acid complexes that exhibit Gaussian-like isotopic distributions and that ranking by ΔX ₀ is effective in these cases only. Users should use alternative approaches , for assessing fit certainty in cases with no or very few silvers.
• 6. Fewer than three resolved isotope peaks in the envelope. Reliable automated charge-state determination requires at least three resolved peaks. Diagnostic: no charge-state annotation appears above the envelope. Action: use the manual charge-state override and treat the assignment as tentative until confirmed at another z.
• 7. Top-ranked Q_clvaries between adjacent values at different charge states. Because surviving candidates after mass filtering typically differ only by one unit in Q _cl, borderline envelope quality can cause an adjacent assignment to rank as best at one z. Diagnostic: top-1 vs top-2 score gap small at one z, large at another z, with disagreement in the best-ranked compositions at different z. Action: inspect the isotope overlay at each z, and select the composition that is consistent across charge states.
• 8. Overlapping isotope envelopes. When two or more species produce isotope envelopes that partially overlap in m/z, NucleoSpec treats the mass spectral peak as a single composition, distorting both the Gaussian-fit centroid and pattern similarity. This situation can occur when byproducts or fragmentation products appear at similar m/z values. Diagnostic: anomalously low pattern similarity despite adequate signal-to-noise; asymmetric or multimodal envelope shape; assigned composition changes when the analysis window is narrowed. Action: visually inspect the envelope for asymmetry. If overlap is suspected, the user may narrow the analysis window to isolate the dominant species (if possible), examine other charge states where the species may be better resolved, or use complementary deconvolution tools before reanalyzing with NucleoSpec. Purification, such as by HPLC, can also reduce the dispersity of products observed in the mass spectrum.

5. Applying NucleoSpec across Types of Nucleic Acid–Silver Complexes

Next, we walk through analysis of representative spectra from each supported sample type, comparing NucleoSpec’s automated output to published manual assignments. The reference compositions in Table were originally determined using manual analysis workflows typical of this field, which commonly involve peak selection, candidate enumeration, isotope pattern generation, and visual comparison of theoretical and experimental envelopes across multiple independent software tools. NucleoSpec recovers these same assignments automatically and rapidly within a single interface, eliminating the time-consuming analysis process and multitool workflow that characterize current practice. Representative cases were chosen to exercise each analysis mode of NucleoSpec end-to-end (sequence and formula input, conjugate handling, and adduct handling). Broader benchmarking on novel systems can be carried out in future application studies.

5.1. DNA-Ag _N with Chlorido Ligands

NucleoSpec can identify DNA-Ag _N that incorporate noncovalent ionic ligands such as chlorido, as demonstrated for the chlorido-protected DNA-Ag₁₆ reported by González-Rosell et al. (Figure a). NucleoSpec accurately assigns molecular formulas to both the chlorido-protected (DNA)₂[Ag₁₆Cl₂] (Q _cl = 8, pattern 1.00 at z = 4^–, 5^–), and a chlorido-free species, (DNA)₂Ag₁₅, that is also resolved in the mass spectrum (Q _cl = 9, pattern similarity 0.996–0.999 at z = 3^–, 4^–) and forms by degradation of the original (DNA)₂[Ag₁₆Cl₂] under ESI. In both cases, the species are correctly assigned as sharing N ₀ = 6 (Figure a), reproducing the compositions reported in the original studies.

3.

Generalization of the analytical framework across nucleic acid–silver complexes. (a) DNA-Ag _N containing two chloridos, with previously reported composition validated by isotope-pattern agreement. (b) SNA-Ag _N with previously reported composition assignment validated by isotope-pattern agreement. (c) Duplex C20–silver complex evaluated under the chemically expected condition N ₀ = 0, yielding a composition that reproduces the experimental isotope envelope. (d) Chemically conjugated system: BCN-modified DNA-Ag _N . Analysis shows the assignment of ${(\mathrm{D}\mathrm{N}\mathrm{A}‐\mathrm{B}\mathrm{C}\mathrm{N})}_2{[{\mathrm{A}\mathrm{g}}_{16}{\mathrm{C}\mathrm{l}}_2]}^{6+}$ , with experimental and theoretical isotope overlays. This demonstrates simultaneous handling of the conjugate, silver cluster, and chloride adduct. The experimental envelope is shifted by ∼2 Da in mass relative to the nominal conjugate mass due to in-source hydrogenation of the BCN alkyne (+2H), consistent with prior work.

5.2. XNA-Ag _N

For XNA-silver complexes such as SNA- and TNA-stabilized nanoclusters, the framework requires only the molecular formula of the XNA strand as input, operating at the level of elemental composition and formal charge rather than nucleotide sequence (Figure b). NucleoSpec correctly assigns (SNA)₂Ag₁₄ and (SNA)₂Ag₁₅ at z = 4^–, and correctly assigns (TNA)₂Ag₈ and (TNA)₂Ag₁₁ at z = 4^–. Pattern similarity scores for the SNA species (0.53 to 0.84) were lower than for DNA-Ag _N systems due to the lower signal-to-noise ratios of the mass spectra; TNA species achieved high similarity (0.92 to 0.96). All ΔX ₀ values remained below 0.08 m/z, confirming correct composition recovery by NucleoSpec.

5.3. Ag­(I)-Mediated Complexes

For nucleic acid–Ag­(I) complexes, compositions can be evaluated under the chemically expected condition that N ₀ = 0 (Q _cl = n _Ag), which can decrease computational time. The platform correctly identified bare duplexes with no silvers (pattern similarity 0.98 to 1.00) and silver-containing duplexes, including (dC20)₂Ag₂₀ (pattern similarity 0.95 at _z = 5^–) and (TNA)₂Ag₈ (pattern similarity 1.00 at z = 3^–) (Figure c).

5.4. Covalent conjugates

A key advantage of NucleoSpec is the ability to easily integrate compositional analysis for chemically modified nucleic acid systems (Figure d). NucleoSpec assigns the composition of a previously reported BCN-modified DNA-Ag _N system as (DNA-BCN)₂[Ag₁₆Cl₂] with N ₀ = 8 (Q _cl = 6) and pattern similarity of 0.98 at z = 4^– and 0.96 at z = 5^–, with centroid errors below 0.09 m/z, reproducing the published species assignment. We note that NucleoSpec’s assignment of N ₀ = 8 matches the findings of Rück et al., because the BCN alkyne undergoes +2H modification during electrospray ionization, adding ∼2 Da to the conjugate mass. As described in Section , item 4, this shifts the automatically assigned N ₀ by +2 and requires the user to apply domain expertise to discern the correct N ₀ for the nanocluster species. When accounting for the 2 additional protons on the BCN groups, the corrected N ₀ = 6 matches the nonconjugated analogue, illustrating that domain expertise remains important for interpreting the results of analysis involving chemical modifications.

NucleoSpec imposes no restrictions on the conjugate formula, which may contain any arbitrary elements, and conjugates need not be present on every strand (n _conj ≤ n _s ). Conjugates are specified through an integrated molecular editor that supports structure drawing, SMILES input, and database lookup, with molecular formulas extracted automatically. This generality enables broad application to emerging conjugate chemistries (including biotin, fluorophores, and other functional group modifications) without algorithmic changes.

5.5. Summary of Performance Across the Example Set

Across the 28 peaks from eight representative samples in Table , the average value of ΔX ₀ is 0.05 ± 0.05 m/z (mean ± SD, median 0.03 m/z). For 25 of the 28 peaks, ΔX ₀ falls below 0.1 m/z (significantly less than one proton-equivalent at the relevant charge states), confirming near-single-proton precision across all the sample types. The largest value of ΔX ₀ (0.22 m/z for (dC20)₂Ag₂₀ at z = 6^–) arises from an asymmetric experimental isotope envelope that shifts the Gaussian fit centroid; the same species achieves ΔX ₀ < 0.07 m/z at both z = 5^– and 7^– and confirms the correct composition. This case illustrates the value of cross-charge state validation: an inflated centroid offset at one charge state is reconciled by consistent assignments at the other charge states. All DNA-Ag _N , adduct, and conjugate samples achieved high confidence in NucleoSpec analysis. TNA-Ag _N and all duplex systems also achieved high confidence (pattern similarity $>0.9$ ). SNA-Ag _N spectra showed mixed confidence: (SNA)₂Ag₁₄ achieved high confidence (0.84), while (SNA)₂Ag₁₅ (0.53) showed moderate confidence due to partially overlapping isotopic envelopes in these experimental data. Even when absolute similarity scores are lower, the correct assignment is reliably identified because |ΔX ₀| resolves adjacent-Q _cl candidates that pattern similarity alone cannot distinguish. Additional validation of NucleoSpec is provided in the Supporting Information, including the 12 peaks from Figure , which are listed in Table S5. An extended validation for four more DNA-Ag _N samples with 63 distinct peaks is also provided in Supporting Information (Section S10); these data include lower signal-to-noise peaks that illustrate the diagnostic steps in Section .

6. Comparison with Manual Workflow and Limitations

6.1. Conventional Manual Workflow

The conventional approach to determine the composition of a nucleic acid–silver complex from its mass spectrum involves manually fitting each mass spectral peak’s isotope envelope using the relationships given above. In the case of DNA-Ag _N , one must also iterate over multiple possible nanocluster core oxidation states to determine Q _cl. The steps involved in this approach are as follows. First, one must determine the charge state z of a resolved product using the spacing of isotope peaks (Δm/z ≈ 1.003/z). Multiplication of the observed m/z peak by z yields the ion mass of the complex. Candidate compositions consistent with this mass are then hypothesized, including the number of nucleic acid strands n _s and the number of total silvers n _Ag. For each candidate mass, a theoretical isotopic distribution is calculated using a tool such as MassLynx and then tuned to match the experimental spectrum at all charge states for which the product is resolved. Because ¹⁰⁷Ag and ¹⁰⁹Ag have nearly equal natural abundances at a ratio of ∼52:48, a DNA-silver complex with at least a few silvers has a characteristic Gaussian-shaped isotope envelope. The centroid of this Gaussian envelope can be used to determine n _s , n _Ag, and Q _cl by comparison to calculated mass distributions. , This is performed by fitting Gaussian functions to calculated mass distributions at a range of Q _cl and selecting the calculated distribution whose centroid is closest to the Gaussian fit centroid of the experimental distribution. This process is laborious and typically requires switching between several software tools to perform various steps in the analysis.

6.2. What Automation Eliminates

NucleoSpec consolidates peak detection, charge-state determination, composition enumeration, isotope-pattern matching, cross-charge state validation, and adduct/conjugate handling into a single browser-based pipeline. By manual analysis, each of these stages was typically carried out in separate general-purpose tools, with the researcher manually transferring candidate compositions between tools. Beyond the bookkeeping burden, the manual workflow introduces three recurring sources of error that domain experts can learn to guard against but that are easy to overlook: (i) charge-state misassignment when adjacent isotope peaks are partially resolved or when a peak of interest overlaps with a less intense peak at a higher charge state (Section , item 1), (ii) violation of the charge-balance constraint N ₀ + N ₊ = n _Ag when candidate compositions are enumerated by mass alone, and (iii) inconsistent assignments across charge states for the same chemical species, as it can be tempting to skip exhaustive cross-checking at multiple charge states because doing so multiplies effort. NucleoSpec is designed to eliminate each of these failure modes: charge-balance is enforced during enumeration rather than checked post hoc, cross-charge state consistency is a one-click operation rather than a time-consuming manual process, and apparent charge-doubling from silver isotopologues is corrected automatically (Supporting Information S1). Table contrasts the manual workflow with NucleoSpec step by step for the case of DNA-Ag _N ESI-MS analysis. Analysis time with the manual workflow varies considerably with analyst experience and spectrum complexity; in the authors’ experience, compositional assignment of a 20-peak DNA-Ag _N spectrum typically requires several hours, compared to a few minutes with NucleoSpec (see Section S6 for per-peak benchmarks).

2. Comparison of Manual ESI-MS Analysis Workflow andNucleoSpec for DNA-Ag _N .

Analysis stage	Manual workflow	NucleoSpec
Peak detection and charge-state determination	Manual peak picking; charge state inferred by eye from isotope spacing	Automated peak detection with correction for apparent charge state doubling due to polyisotopic silver
Composition enumeration	Candidates generated one-by-one using external isotope calculator	Exhaustive enumeration over (n s , n Ag, Q cl) filtered by N 0 + N + = n Ag
Isotope-pattern matching	Theoretical isotope envelopes are generated from user-defined compositional formulas	IsoSpecPy envelopes ranked by mean (cosine, Pearson) and centroid X 0 offset
Cross-charge state validation	Manual repetition of analysis at each charge state	One-click re-evaluation of the product at all observed z, with charge state flagging
Adduct/conjugate handling	Additional mass/charge added manually for each modification	Conjugate-aware mass model; adducts as charge-bearing mass offsets
Typical time per 20-peak spectrum	Hours (analyst-dependent)	Minutes

To summarize the degree of automation in each stage: peak detection, charge state determination (including automatic correction of apparent charge doubling), composition enumeration with charge-balance enforcement, isotope pattern generation, and pattern matching are fully automated and require no user input beyond the initial spectrum upload and sequence entry. Cross-charge state validation and adduct/conjugate library configuration are simplified to single-click operations that previously required repetitive manual effort across multiple tools. Several tasks still require user expertise: overriding the charge state for low signal-to-noise envelopes where automated detection is unreliable, recognizing and handling overlapping isotope envelopes from co-ionizing species (Section ), reconciling in-source chemical modifications that shift the apparent N ₀ (e.g., the BCN + 2H modification), and interpreting borderline pattern similarity scores or conflicting cross-charge state assignments.

6.3. Limitations and Caveats

Several limitations of the current NucleoSpec framework should be noted and originate from the fact that automated analysis is not a cure for poor quality data. First, overlapping isotope envelopes from multiple species at similar m/z values cannot be deconvolved; the algorithm assumes each detected envelope corresponds to a single composition. This scenario is particularly relevant when purity of a nucleic acid–silver complex is limited (see Section for diagnostic guidance). Second, pattern similarity degrades at low signal-to-noise ratios, as demonstrated for some SNA cases in Table , where scores fell below 0.5 despite correct composition recovery. Third, reliable charge state determination requires at least three resolved isotope peaks, precluding analysis of poorly resolved or broad features; for low signal-to-noise peaks where the autodetected charge state is unreliable, the user can manually override the charge assignment in the interface, as illustrated for two peaks in the full-spectrum benchmark (Supporting Information S9). Fourth, XNA complexes require longer computation times (∼5–7 s per peak) compared with DNA-Ag _N (∼0.6–0.9 s per peak, increasing to ∼1.3–1.5 s per peak for conjugate-modified DNA-Ag _N ) and DNA-Ag­(I) complexes (∼0.2 s per peak) due to formula-based enumeration over a broader composition space. Large silver clusters (n _Ag > 25) will also increase computation time due to broader isotope manifolds. Fifth, in-source chemical modifications (such as the +2H modification of the BCN alkyne reported by Rück et al.) can shift the apparent N ₀ value, requiring domain expertise to reconcile algorithmic output with chemical expectation. Sixth, the composition search space is defined by the user-specified nucleic acid sequence, integer numbers of silvers, and optionally enabled adduct and conjugate libraries. Species with compositions outside this search space are not matched in order to reduce analysis time for users. In such cases, NucleoSpec will either return no candidate within the mass tolerance or candidates with anomalously large centroid offsets and low pattern similarity scores, both of which serve as diagnostic indicators that the observed species may lie outside the current search space. Even when no candidate passes the mass tolerance filter, NucleoSpec displays the closest approximate match to enable the user to compute the mass residual and hypothesize an unknown adduct or modification. Users encountering unassigned peaks or uniformly poor-scoring candidates can expand the adduct library with custom entries to include additional conjugate modifications or perform manual analysis to identify the unknown species (see Supporting Information section S10 for examples of automated Na⁺ and Cl^– adduct detection when bare (no-adduct) compositions do not yield satisfactory centroid matches).

Finally, it should be noted that all benchmark spectra in this study were drawn from published data of sufficient quality to be reported in the literature. The assignments for DNA-Ag _N , chloride-adducted DNA-Ag _N , and BCN-conjugate shown here were performed after their original publication: the NucleoSpec assignments were obtained for spectra reported in previous studies but without prior knowledge of the expected results. The XNA-Ag _N and Ag­(I)-mediated nucleic acid duplex assignments are partially concurrent: an early version of NucleoSpec was used during the original compositional analysis published in that work and confirmed by manual workflow. Further application of NucleoSpec to genuinely novel systems will test the framework’s generality. As noted above, NucleoSpec is not a solution for overcoming poor-quality data, and users are encouraged to follow best practices for experimental ESI-MS at near-native conditions for resolving nucleic acid-silver complexes. This is a challenge that also applies equally to conventional manual workflows. NucleoSpec is intended to expedite mass spectral analysis for these emerging nucleic acid materials.

7. Conclusions

NucleoSpec consolidates ESI-MS compositional analysis of nucleic acid–silver complexes into a single browser-based pipeline that is designed to enforce the charge-balance constraint N ₀ + N ₊ = n _Ag and to support cross-charge state validation, conjugate-aware mass modeling, and adduct handling within a unified interface. Validation across DNA-Ag _N , XNA-Ag _N , Ag­(I)-mediated duplexes, and covalently conjugated DNA-Ag _N reproduces published manual assignments to near-single-proton precision, including spectra with low signal-to-noise ratios. A 20-peak DNA-Ag _N spectrum is analyzed in minutes using NucleoSpec rather than the hours typically required when iterating between independent tools. Thus, this automated workflow removes one of the two key experimental bottlenecks in high-throughput discovery of nucleic acid–silver complexes. The source code is freely available under CC BY-NC 4.0 at https://github.com/copplab/NucleoSpec, with a hosted instance at https://huggingface.co/spaces/copplab/NucleoSpec for browser-based use without installation.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmeasuresciau.6c00178.

• Algorithmic details for charge state determination, composition enumeration, isotope-pattern generation and similarity metrics, cross-charge state validation, and runtime benchmarks (sections S1–S8); full-spectrum validation against published assignments for one DNA-Ag _N system (section S9); extended validation across four additional DNA template sequences totaling 63 peaks (section S10); interface screenshot (Figure S1); isotope-pattern overlay comparisons (Figure S2); monoisotopic masses, adduct library, runtime benchmarks, candidate ranking selectivity, and per-system validation tables (Tables S1–S9) (PDF)

The authors declare no competing financial interest.