acceptodds
Withdrawn from ICLR 2027

Transferable Mass Spectrum Prediction via Reference-Guided Test-time Specialization

Withdrawn from ICLR 2027, so left out of the paper lists and the map.

17 views

Abstract

Tandem mass spectrum prediction supports compound identification across metabolomics, natural-product discovery, and environmental analysis. However, pretrained predictors often degrade under shifts in chemical space and acquisition conditions, while retraining domain-specific models from scratch is costly. We introduce SPARC, a retrieval-guided test-time specialization framework that adapts a pretrained predictor using a spectral reference library without accessing test-query spectra. For each target query, SPARC retrieves chemically related reference spectra to recalibrate fragment intensities within the learned fragmentation space. SPARC combines reference-guided spectral adaptation with reliability-aware consistency, using reconstruction behavior on retrieved spectra to selectively preserve trustworthy predictions during continual specialization. Across MassSpecGym, NPLIB1 and application-specific GNPS libraries, SPARC improves spectral prediction under multiple transfer settings. These results establish retrieval-guided test-time specialization as a practical strategy for extending pretrained MS/MS predictors to specific chemical and acquisition domains, with continual test-time training providing further refinement during deployment.

Primary area:
applications to physical sciences (physics, chemistry, biology, etc.)
Keywords:
Machine Learning, Tandem Mass Spectrometry, Mass Spectrum Prediction, Test-time training, Domain Adaptation

Reviews

Among:
Shape:
1234this paper, Bottom 27%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 2.00 over 3 reviews, above 27% of the others.
Reviewer_4U5kOpenReview ↗
Rating
1
Confidence
4
Informativeness
2
Presentation
3
Soundness
1
Read review
Paper summary

This paper proposes SPARC, which adapts a pretrained MS/MS predictor (Iceberg) to a target domain one query at a time. For each query, it retrieves 64 adduct- and instrument-matched labeled spectra based on fingerprint similarity and uses them to fine-tune the intensity model rather than using the query’s measured spectrum. The fragment generator is frozen, and the intensity model is updated with a similarity-weighted reconstruction loss and a teacher-student consistency loss. An EMA teacher, stochastic restoration, and rollback are used to limit model drift. The experiments cover adduct specialization, cross-library transfer, within-domain refinement, and five GNPS libraries.

Critical strengths

SPARC aims to adapt a pretrained MS/MS predictor to a target dataset at the prediction time. It shows improvement from the pretrained predictor without adaption.

Critical weaknesses

My main concern is that the experiments do not clearly separate the benefit of SPARC from simply training or fine-tuning Iceberg on the same labeled spectra available to SPARC. In the GNPS experiments, these spectra are also part of Iceberg’s original training data, while on NPLIB1, the only standard fine-tuning baseline performs worse than training from scratch. Moreover, although SPARC substantially improves over the pretrained model in the transfer settings, its margins over models trained on the target data are generally small, as are its gains within domain and on several GNPS libraries. These results are also reported without variance estimates, making it difficult to assess their robustness. Overall, the improvements appear to come primarily from increased peak precision rather than a consistent improvement across evaluation metrics.

Weakness 1: The baselines do not adequately isolate the benefit of SPARC, and several appear under-tuned.

SPARC has access to labeled target spectra in every experiment, but the comparisons do not consistently control for this advantage. For MassSpecGym and NPLIB1, the support library is the target training split, while validation labels are used for checkpoint selection (Sec. 4). For GNPS, SPARC uses the MSG training split, which is also the data used to train the Iceberg baseline in Table 4. Thus, SPARC is effectively performing additional target-data fitting, rather than adapting with information unavailable to the baseline.

The within-domain NPLIB1 result illustrates this issue. Re-fitting Iceberg on NPLIB1 increases CosSim from 0.586 to 0.616 (Table 2), even though there is no domain shift. This suggests that part of SPARC’s improvement may come from local per-query fitting, similar to test-time training on nearest neighbors [1], which is not evaluated.

The margins over models trained directly on target data are also modest. On MSG to MSG-MNa, SPARC(val) improves over Iceberg trained from scratch by only 0.022 CosSim (0.370 vs. 0.348; Table 1), while removing retrieval reduces SPARC to 0.344 (Table 5). On NPLIB1, retrieval-weighted fine-tuning alone reaches 0.587, essentially matching Iceberg trained on NPLIB1 (0.586; Sec. 5.2). Meanwhile, global fine-tuning reaches only 0.5545, below training from scratch, raising questions about whether this baseline was adequately tuned. NEIMS and GrAFF-MS also perform substantially worse than Iceberg, which makes those comparisons difficult to interpret.

Two simple baselines are particularly important: Iceberg pretrained on the union of the available libraries, with and without standard fine-tuning on the target split, and a similarity-weighted average of the 64 retrieved spectra. TENT and CoTTA use no labels, so comparisons with them do not isolate the contribution of retrieval.

Weakness 2: The improvements are small, inconsistent across metrics, and reported without variance estimates.

Several reported gains are quite small. On MSG M+H, SPARC(val) increases CosSim from 0.529 to 0.532, while MSE worsens from 0.853 to 0.907 (Table 2). On GNPS, the CosSim gain is only 0.005 for 3-Hydroxy Acyl Amides, and SPARC improves all three metrics on only two of five libraries (Table 4 and Sec. 5.4). The online TTT stage contributes only 0.006–0.020 CosSim (Tables 1 and 2).

No results are reported across multiple seeds or with standard deviations. This is important because SPARC is stochastic and order-dependent (Sec. 6), making it difficult to determine whether the 0.006–0.031 differences between ablations in Table 5 are robust.

There is also a substantial precision-recall trade-off. Peak recall drops from 0.550 to 0.449 on M+H and from 0.159 to 0.097 on M+Na (Table 3), but this is not discussed despite the potential impact of missing peaks on downstream identification.

Finally, the evaluation protocol makes comparison with prior work difficult. The paper retains only the top 100 predicted bins, evaluates only spectra that all methods can predict, filters support spectra by scaffold, and splits MassSpecGym by adduct (Sec. 4). The selection of the five GNPS libraries is also not explained. Results under the official MassSpecGym protocol and a downstream molecule-retrieval experiment would provide stronger evidence that the reported spectral improvements translate into better identification.

Weakness 3: The methodological novelty and generality are limited, and the computational cost is unclear.

Several components build directly on existing methods. The EMA teacher and stochastic restoration come from CoTTA, while retrieval-weighted fine-tuning is a form of local learning (Sec. 3.4). The main new component, support-calibrated bin-wise weighting (Eqs. 5–6), contributes 0.024 CosSim in one unreplicated experiment (Table 5).

SPARC is evaluated only with Iceberg, so its applicability to other pretrained MS/MS predictors remains untested. Because the fragment generator is frozen, SPARC also cannot recover diagnostic fragments missing from the pretrained model (Sec. 6). The entropy guard further favors sharper spectra by rejecting updates that increase entropy, which seems somewhat inconsistent with the paper’s argument that lower entropy does not necessarily indicate better fragmentation.

The computational overhead is also important but unreported. Each query requires processing 64 retrieved spectra with forward and backward passes, rather than a single forward pass. This could be substantial for large-scale in-silico library generation. In addition, the model state carries across queries, so predictions depend on query order (Sec. 6). The paper should quantify both the computational cost and the magnitude of this order dependence.

References

[1] Hardt, Moritz, and Yu Sun. "Test-time training on nearest neighbors for large language models." International Conference on Learning Representations. Vol. 2024. 2024.

Questions for authors

How does SPCAC compare with the test-time training on nearest neighbors [1] or the direct fine-tuning of iceberg on the target dataset? How does SPARC perform on the official MassSpecGym protocol? What is the contribution of the contribution of retrieval to the improvement by SPARC?

[1] Hardt, Moritz, and Yu Sun. "Test-time training on nearest neighbors for large language models." International Conference on Learning Representations. Vol. 2024. 2024.

Reviewer_bcbbOpenReview ↗
Rating
2
Confidence
2
Informativeness
3
Presentation
2
Soundness
2
Read review
Paper summary
  • This paper adapts a retrained Iceberg spectrum predictor to each query at test time. It freezes the fragment generator and fitunes only the intensity scorer using 3 comments: supervision from retrieved neighbors, per-bin brake and drift control. Experiments show the effectiveness of the proposed method.
Critical strengths

-The proposed method freezes the discrete fragment generator and adapt only continuous intensities which is well motivated decomposition and domain-aware design: reallocates intensity among exisitng candidates.

  • Using the teacher's measure error on labeled neighbors to decide where to trust is a reasonable idea. It is better motivated than the confidence heuristics used in generic test-time adaption.
Critical weaknesses
  • Limited novelty in the components. The EMA teacher, stochastic restoration and consistency regularization come from CoTTA and mean-teacher method.
  • Fine-tuning on retrieved nearest neighbors at test time has been proposed in local learning (Bottou & Vapnik 1992) and test-time training on nearest neighbors (ICLR 2024). A discussion on these should be given.
  • Many claimed improvements are insignificant. Many claimed improvements are small. In the ablation, removing stochastic restoration costs about 0.006 cosine and removing rollback about 0.012. Several Table 2 and Table 4 differences are in the second decimal.
  • The ablation is run on a single transfer (MSG→M+Na) only.
Questions for authors

see weaknesses.

Reviewer_qAC7OpenReview ↗
Rating
3
Confidence
1
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

SPARC adapts a pretrained molecular spectrum predictor using measured reference spectra that are chemically related and compatible with the target acquisition setting. It keeps fragment generation fixed and updates only the intensity model, using support reconstruction and support-error-weighted teacher consistency. The method also uses EMA, restoration, and rollback for stable adaptation. Experiments cover adduct and library transfer, validation-based specialization followed by online continuation, and five GNPS application libraries without target validation.

Critical strengths
  • The motivation is clear. The paper addresses the practical problem of adapting spectrum predictors to specific chemical and experimental domains. Using existing reference spectra provides a useful source of supervision when measured spectra for the query molecules are unavailable.
  • The experiments also include repeated runs with shuffled query orders, which helps assess the stability of the method under different online adaptation sequences.
Critical weaknesses
  • The paper does not compare against MARASON, a closely related method that also uses measured reference spectra to improve an ICEBERG-based predictor.
  • The evaluation does not show whether better spectrum prediction leads to better molecular identification. AP and NDCG measure peak-level prediction quality, while the downstream task is to rank candidate molecules. Higher spectral similarity does not necessarily mean that the correct molecule is easier to distinguish from incorrect candidates.
Questions for authors
  1. Could you compare SPARC with MARASON and retrieved fine-tuning under the same source-training split, reference library, target-validation access, and test set?
  2. Could you evaluate molecular retrieval using the official formula-based candidate lists and report Hit@1, Hit@5, and Hit@20?
  3. For MSG->M+Na and MSG->NPLIB1, could you evaluate target-validation budgets of (N = 0, 25, 100), and all available validation molecules?
  4. Could you evaluate (K = 1, 8, 16,) and (64) references, and compare structure-only, acquisition-only, and combined retrieval?

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Discussion (0)

Sign in to comment.