ProtoPol: Diverse Research Artifacts

• ProtoPol is an overloaded term in research, referring to distinct methods in process discovery, biochemical protocol modeling, spectro-polarimetric instrumentation, and protocol programming.
• In process mining, ProtoPol incrementally selects representative traces using clustering and conformance metrics to achieve improved model fitness and precision.
• In experimental biology and astronomy, ProtoPol underpins a probabilistic protocol language and a dual-beam spectro-polarimeter with dedicated data reduction pipelines for precise analysis.

Searching arXiv for “ProtoPol” and related variants to disambiguate the term across research contexts. ProtoPol is a label used for several unrelated research artifacts rather than a single unified method. In the cited literature, it refers to an incremental prototype-selection method for process discovery, a probabilistic language for modeling and optimizing experimental biochemical protocols, a medium-resolution echelle spectro-polarimeter for the PRL telescopes at Mt Abu, and a programming paradigm centered on protocol-mediated object connection. The shared name obscures substantial differences in domain, formalism, and evaluation, ranging from conformance-guided event-log reduction to Gaussian semantics for laboratory operations, dual-beam spectro-polarimetric data reduction, and communication-layer software architecture (Sani et al., 2019, Cardelli et al., 2021, Maiti et al., 26 Mar 2026, Chen et al., 2015).

1. Scope of the term

The term appears in multiple technical contexts. The following usages are directly attested in the literature.

Usage of ProtoPol	Domain	Defining characterization
Incremental prototype selection	Process mining	Representative traces are iteratively selected from event logs for process discovery
Protocol-modeling language	Experimental biology	Protocols are programs over chemically meaningful sample states with uncertainty
Medium-resolution spectro-polarimeter	Astronomical instrumentation	Dual-beam echelle spectro-polarimetry over the visible band
Protocol programming	Software systems	Protocols are first-class constructs for connecting objects and coordinating communication

Chronologically, the papers on arXiv place protocol programming in 2015, process-discovery ProtoPol in 2019, the biochemical protocol language in 2021, and the Mt Abu instrument paper in 2026. This suggests that “ProtoPol” functions as an overloaded term whose meaning must be resolved by disciplinary context rather than by name alone (Sani et al., 2019, Cardelli et al., 2021, Maiti et al., 26 Mar 2026, Chen et al., 2015).

2. ProtoPol in process discovery

In process mining, ProtoPol is an incremental prototype selection method for process discovery. Its purpose is to avoid discovering a model from the whole event log directly; instead, it selects a small, representative set of traces, discovers a process model from those prototypes, evaluates conformance on the original log, and iteratively augments the prototype set with representatives of deviating behavior until the quality trade-off no longer improves (Sani et al., 2019).

The workflow has four major steps. First, traces are clustered using edit distance on trace sequences; the paper uses K-Medoids because it returns actual traces as representatives. Second, the centroid trace of each cluster is selected as a prototype and a process discovery algorithm is run on the prototype sublog. Third, the discovered model is assessed on the whole original log using conformance metrics such as fitness and precision, combined into $F$-measure or $F_\beta$-measure. Fourth, traces that do not fit perfectly are identified, reclustered, and their representatives are added as new prototypes. The loop stops when the current $F_\beta$ does not improve over the previous iteration.

The paper formalizes the event log as

$L \in M(\mathcal{U}_A^*)$

and a clustering of $L$ into $n$ clusters as

$\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$

with disjoint clusters whose multiset union reconstructs the original log. The balancing criterion is the standard

$$F_{\beta} = (1 + \beta^2 ) \cdot \frac{Precision \cdot Fitness}{(\beta^2 \cdot Precision)+Fitness}.$$

The stopping rule follows the observed iteration behavior: as more prototypes are added, fitness usually increases, precision usually decreases, and complexity tends to increase, so a weighted trade-off score is used rather than a monotone inclusion rule.

ProtoPol is implemented as a Prototype Selection plug-in in ProM and was ported to RapidProM. The implementation allows users to choose the number of clusters or prototypes, the $\beta$ value, and the process discovery algorithm. In the experiments, the method is used with Inductive Miner, ILP Miner, and Split Miner, and is evaluated on eight real event logs: BPIC 2012, BPIC 2018 – Department, BPIC 2018 – Inspection, BPIC 2018 – Reference, BPIC 2019, Hospital, Road, and Sepsis.

The reported empirical effect is a better balance of fitness and precision than no preprocessing, Sampling, or Statistical subset selection, often with reduced model complexity. On the Road log, one comparison reported Inductive Miner on the full log at fitness $0.78$ and precision $F_\beta$0, whereas ProtoPol with Inductive Miner reached fitness $F_\beta$1 and precision $F_\beta$2. The paper also notes limitations: the method may converge to a local optimum, depends on the number of clusters, $F_\beta$3, and discovery algorithm, and assumes the common trade-off in which adding traces tends to improve fitness but hurt precision (Sani et al., 2019).

3. ProtoPol as a language for biochemical protocols

In experimental biology, ProtoPol is a protocol-modeling language for describing biochemical or biological protocols together with the underlying reaction-network dynamics. Its central purpose is to unify a mechanistic model of the molecular system with the laboratory operations used to manipulate samples, collect data, and optimize protocol parameters. In this formulation, a protocol is not merely a text procedure but a program executed over sample states that carry concentrations, covariance, volume, and temperature (Cardelli et al., 2021).

A sample is modeled as a state of a Chemical Reaction System: $F_\beta$4 where $F_\beta$5 is the set of chemical species, $F_\beta$6 the set of reactions, $F_\beta$7 the concentration mean vector, $F_\beta$8 the covariance matrix, $F_\beta$9 the sample volume, and $F_\beta$0 the temperature. The underlying chemistry is a CRN $F_\beta$1, and the time evolution uses a Gaussian or Linear Noise Approximation semantics with flux

$F_\beta$2

and diffusion term

$F_\beta$3

The protocol syntax is intentionally small. A protocol may be a sample literal, a let-binding, Mix, Split, Equilibrate, or Dispose. The semantic interest lies in how these operations transform Gaussian sample states. For example, mixing two samples produces

$F_\beta$4

while equilibration advances the sample according to the CRN dynamics: $F_\beta$5 The paper distinguishes extrinsic noise from intrinsic noise and adopts intrinsic-noise semantics, under which stochastic chemistry in each sample is modeled directly and splitting produces uncorrelated samples under the well-mixed assumption.

ProtoPol is also designed for data integration and optimization. The paper introduces an optimization directive of the form

$F_\beta$6

and treats protocol outputs as a Gaussian-process prior updated by experimental observations. The posterior mean and covariance follow standard GP regression formulas, and expected-cost minimization is carried out with a reparameterized gradient estimator. The stated objective is

$F_\beta$7

The case studies include a toy CRN optimization, a split-and-mix workflow, and a simplified Gibson assembly protocol. In the Gibson assembly example, the optimization variables are the initial concentration $F_\beta$8 and reaction time $F_\beta$9, and the cost trades off the desired product concentration, deviation from a reference time, and deviation from a reference concentration. The paper’s broader claim is that ProtoPol differs both from protocol languages that treat samples as black boxes and from CRN-only models that do not represent laboratory operations: it provides a single formal framework for reaction kinetics, protocol execution, uncertainty propagation, data integration, and protocol optimization (Cardelli et al., 2021).

4. ProtoPol as a spectro-polarimeter

In astronomical instrumentation, ProtoPol is a medium-resolution dual-beam echelle spectro-polarimeter developed for the PRL 1.2m and 2.5m telescopes at Mt Abu, India. The cited paper is the second and final part of the instrument’s development series and concentrates on the dedicated data-reduction pipeline, on-sky characterization, performance evaluation, and first science results. ProtoPol was built entirely from commercially available off-the-shelf components and was originally conceived as a prototype for a larger multi-mode instrument, M-FOSC-EP, before being promoted to a backend instrument after laboratory and on-sky validation (Maiti et al., 26 Mar 2026).

The instrument covers approximately $L \in M(\mathcal{U}_A^*)$0–$L \in M(\mathcal{U}_A^*)$1 Å, with the Blue cross-disperser spanning about $L \in M(\mathcal{U}_A^*)$2–$L \in M(\mathcal{U}_A^*)$3 Å and the Red cross-disperser about $L \in M(\mathcal{U}_A^*)$4–$L \in M(\mathcal{U}_A^*)$5 Å. Across the orders, the measured spectral resolution is about $L \in M(\mathcal{U}_A^*)$6–$L \in M(\mathcal{U}_A^*)$7 Å for the Blue arm and $L \in M(\mathcal{U}_A^*)$8–$L \in M(\mathcal{U}_A^*)$9 Å for the Red arm, summarized as roughly $L$0–$L$1 Å overall. On the PRL 2.5m telescope, the paper reports that an SNR of $L$2 is achieved for an $L$3 source in $L$4 hour of integration time, with throughput estimated to be about $L$5 including atmospheric transmission, telescope reflectivity, optics, and CCD efficiency. Linear polarization accuracy is reported as $L$6–$L$7 in $L$8 hours for a source with $L$9, and instrumental polarization is around $n$0.

A major contribution is the Python 3.11 reduction pipeline using Astropy, NumPy, and SciPy. The pipeline performs bias and dark subtraction, cosmic-ray removal using lacosmic, order tracing, scattered-light subtraction, extraction of ordinary and extraordinary beams, wavelength calibration using Uranium-Argon lamps, sky subtraction from separately observed sky frames, alignment of spectra from the four half-wave-plate positions, computation of Stokes parameters, order joining, relative flux calibration, dynamic binning, and optional user-guided continuum refinement. The linear spectro-polarimetric reductions use the standard dual-beam formalism: $n$1

$n$2

with

$n$3

The on-sky characterization used unpolarized standards, polarized standards, Jupiter for throughput estimation, and faint M-dwarfs for spectroscopic limiting magnitude. The paper reports that no significant anomalous polarization across strong absorption lines is seen in unpolarized standards, though a small pseudo-periodic ripple with a $n$4–$n$5 Å period appears in some intervals and is attributed to polarization fringes. Polarized standards follow the Serkowski law well, with an average deviation of about $n$6 over $n$7–$n$8 Å.

The first science observations include Herbig Ae/Be stars, classical Be stars, symbiotic stars, and AGB/post-AGB stars. Reported phenomena include no line effect in GU CMa and MWC 442, depolarization in 27 CMa and Il Cep, intrinsic line polarization in MWC 158 across H$n$9, H$\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$0, and H$\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$1, McLean-type signatures in MWC 480 and MWC 758, polarized Raman bands in AG Dra and Z And, and continuum polarization trends in AGB/post-AGB stars. In this usage, ProtoPol denotes an instrument and pipeline, not an algorithmic framework or protocol language (Maiti et al., 26 Mar 2026).

5. ProtoPol as protocol programming

In systems research, ProtoPol denotes protocol programming, a programming paradigm whose central claim is that conventional programming encapsulates attributes and behaviors into objects but does not provide a native mechanism to connect those objects and govern communication among them. The paradigm introduces protocols as predefined communication patterns that connect objects and coordinate interactions, with the stated goal of supporting concurrency, parallelism, distribution, pipeline organization, and adaptability (Chen et al., 2015).

The proposed implementation uses a layered architecture. The transmission layer builds the infrastructure for message transmission and establishes the connecting network. The service layer supports service registration, request, and execution, and supervises services provided by the domain layer. The domain layer contains application-specific logic implemented using traditional techniques. This separation is intended to preserve object autonomy while externalizing communication concerns into the protocol layer.

The paper defines a generic message class as:

$\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$7

The transmission layer is implemented by a class called PTrans, and each instantiated object creates three concurrent threads: configuring, connecting, and acting. The paper defines five transmission-layer actions: CONFIG, HELLO, ECHO, TICK, and EXIT. The connection network uses a tree topology, with a Master as root. The initialization protocol proceeds by child creation, exchange of connecting handles through CONFIG, broadcast of HELLO, and recursive ECHO responses once subtrees are ready. The paper also notes timeout-triggered restart if some thread is blocked or suspended.

The service layer is implemented by PService, which inherits from PTrans, so every service object is automatically connected into the network. The service-layer actions are REG, REQ, and ACK. Service descriptions are stored in PServicedepict, recording service identification, location, execution mode, reentrancy, and routing information. Registration proceeds from the Master downward and back upward so that the Master eventually learns all services in the system. Requests can be synchronous or asynchronous, are matched via string matching, and may be executed by thread-based or function-based services. If code is reentrant, the system may create a new thread per request, dynamically schedule requests, and perform load balancing.

The evaluation uses three toy services with workloads of $\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$2, $\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$3, and $\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$4 FMI, instantiated as three objects plus one Master. Four execution styles are compared: monotone, sequence, pipeline, and super pipeline. The reported result is qualitative rather than theorem-level performance analysis: the sequence model is slower than the monotone model because of message-transmission overhead; the pipeline model offsets communication cost by overlapping computation; and the super pipeline achieves the highest performance due to load balancing. When the workload is reduced to $\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$5 FMI, message cost dominates and speedup becomes less than $\xi(L,n)= \{ l_1, l_2, \dots, l_n \},$6. The paper therefore presents ProtoPol as an architectural prototype centered on connection and communication rather than as a mature language or formally verified runtime (Chen et al., 2015).

6. Related and commonly confused terms

Because the label is overloaded, ProtoPol is easily conflated with other similarly named systems. The literature distinguishes it from several nearby terms. TinyProto is a communication-efficient prototype-based federated learning framework that exchanges sparse class prototypes rather than model parameters; its core mechanisms are Class-wise Prototype Sparsification and adaptive prototype scaling (Lee et al., 6 Jul 2025). ProtoPilot is a self-evolving multi-agent system for automated generation and execution of biological protocols, evaluated on 294 synthetic-biology and molecular-biology tasks and on real wet-lab runs (Jiang et al., 30 Jun 2026). ProTo is the Program-guided Transformer for program-guided tasks, with semantic and structural program guidance implemented through cross-attention and masked self-attention (Zhao et al., 2021). PROTOCOL, spelled ProtoCol in the paper’s details, is a late-interaction retrieval model for remote protein homology search that uses residue-level embeddings and ColBERT-style MaxSim scoring (Cohn et al., 27 May 2026). ProtoT, the Prototype Transformer, is an autoregressive language-model architecture that replaces self-attention with a prototype-based mixer designed for interpretability (Yordanov et al., 12 Feb 2026).

A common misconception is therefore to treat ProtoPol as a stable acronym with one accepted expansion. The cited record does not support that reading. Depending on the field, it denotes prototype-based event-log reduction, protocol semantics for biochemical experimentation, an astronomical polarimeter, or a protocol-centric programming paradigm. A plausible implication is that any technical discussion of ProtoPol should resolve the disciplinary context explicitly before comparing methods, equations, software, or empirical results (Lee et al., 6 Jul 2025, Jiang et al., 30 Jun 2026, Zhao et al., 2021, Cohn et al., 27 May 2026, Yordanov et al., 12 Feb 2026).