Mamba-SSM with LLM Reasoning for Feature Selection: Faithfulness-Aware Biomarker Discovery

Pushpa Kumar Balan; Aijing Feng

q-bio.QMApr 15, 2026

Mamba-SSM with LLM Reasoning for Feature Selection: Faithfulness-Aware Biomarker Discovery

Authors: Pushpa Kumar Balan, Aijing Feng

Organizations: Department of Computer Science and Cybersecurity, University of Central Missouri

Abstract

Gradient saliency from deep sequence models surfaces candidate biomarkers efficiently, but the resulting gene lists can be contaminated by tissue-composition confounders that degrade downstream classifiers. We study whether LLM chain-of-thought (CoT) reasoning can filter these confounders, and whether reasoning quality is associated with downstream performance. We train a Mamba SSM on TCGA-BRCA RNA-seq and extract the top-50 genes by gradient saliency; DeepSeek-R1 evaluates every candidate with structured CoT to produce a final 17-gene set. On the held-out test split, the raw 50-gene saliency set (no LLM) performs worse than a 5,000-gene variance baseline (AUC 0.832 vs. 0.903), while the LLM-filtered set surpasses it (AUC 0.927), using 294x fewer features. A faithfulness audit (COSMIC CGC, OncoKB, PAM50) shows that 6 of 17 selected genes (35.3%) are validated BRCA biomarkers, while 10 of 16 known BRCA genes present in the input were missed - including FOXA1. This divergence between downstream performance and reasoning faithfulness suggests selective faithfulness in this setting: targeted confounder removal can improve predictive performance without comprehensive recall.

Figures & tables

Figure 1: End-to-end neuro-symbolic pipeline. High-dimensional TCGA-BRCA RNA-seq is processed by Mamba SSM to produce gradient saliency scores (top-50 candidate genes), then filtered by structured DeepSeek-R1 CoT reasoning to form a 17-gene BRCA-specific subset used in downstream comparison.

Method	Genes	Accuracy	F1	AUC
B1 Variance baseline	5,000	0.8785	0.8941	0.903
B2 Mamba saliency only (no LLM)	50	0.7247	0.7813	0.832
B3 Mamba + LLM structured CoT	17	0.8907	0.9033	0.927

Table 1: Classification performance on the held-out test set (20% of 1,231 samples). All conditions use the same Mamba architecture and training procedure; only the input gene set changes.

Metric	Value
Selected genes with validated BRCA evidence	6/17 (35.3%)
Known non-BRCA genes incorrectly kept	3/17 (17.6%)
Genes with no ground-truth label (unverifiable)	8/17 (47.1%)
Known BRCA genes available in top-50 input	16
Correctly kept by LLM	6
Missed by LLM (false negatives)	10

Table 2: Faithfulness of LLM-selected gene set against curated BRCA ground truth (COSMIC CGC + OncoKB + PAM50,

N=101

validated genes). The top-50 saliency input contained 16 validated BRCA genes.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 2: Performance Comparison: Accuracy, F1, and AUC metrics across experimental conditions. B3 utilizes 250x fewer genes than B1 but achieves higher predictive stability.

Figure 3: Raw Gradient Saliency Heatmap for the top-50 genes in TCGA-BRCA samples, highlighting the noisy feature space prior to LLM intervention.

Figure 4: Visualization of the Agentic Chain-of-Thought (CoT) process, mapping Mamba saliency to biological rationale.

Feature Selection

Feature selection identifies compact, informative variables or input dimensions to improve prediction, efficiency, interpretability, and statistical inference.

See all 196 papers

PiCSAR: Probabilistic Confidence Selection And Ranking for Reasoning Chains
Aug 29, 2025Joshua Ong Jun Leang, Zheng Zhao, Aryo Pradipta Gema +7Confidence Estimation in Language Models Best-of-N Sampling
Imperial College London · University of Edinburgh · UCL
When Do Biological Reasoning Models Use Their Biological Inputs?
Oct 1, 2026Ada Fang, Nikitha Thoduguli, Lukas Fesser +3Faithfulness of Language Model Explanations Causal Interventions in Language Models
Harvard University · Massachusetts Institute of Technology
ROOFS: RObust biOmarker Feature Selection
Jan 8, 2026Anastasiia Bakhmach, Paul Dufossé, Simon Charpigny +4Scientific Discovery Clinical Prediction
1*COMPO Team (COMPutational Pharmacology and clinical Oncology), Inria Sophia Antipolis – Méditerranée, Inserm U1068, CNRS UMR7258, Institut Paoli-Calmettes, Aix-Marseille University, Marseille, France. · 2Veracyte SAS, Marseille, France. · 3Assistance Publique-Hôpitaux de Marseille (APHM), Marseille, France. +1

Aug 29, 2025J/K move · Enter open · S save

PiCSAR: Probabilistic Confidence Selection And Ranking for Reasoning Chains

Joshua Ong Jun Leang, Zheng Zhao, Aryo Pradipta Gema, Sohee Yang+6

Imperial College London · University of Edinburgh · UCL

Confidence Estimation in Language Models Best-of-N Sampling LLM Reranking LLM Uncertainty Estimation Algorithmic Reasoning LLM Reasoning

Open paper AskPDF arXiv

Best-of-n sampling improves the accuracy of large language models (LLMs) and large reasoning models (LRMs) by generating multiple candidate solutions and selecting the one with the highest reward. The key challenge for reasoning tasks is designing a scoring function that can identify correct reasoning chains without access to ground-truth answers. We propose Probabilistic Confidence Selection And Ranking (PiCSAR): a simple, training-free method that scores each candidate generation using the joint log-likelihood of the reasoning and final answer. The joint log-likelihood of the reasoning and final answer naturally decomposes into reasoning confidence and answer confidence. PiCSAR achieves substantial gains across diverse benchmarks (+10.18 on MATH500, +9.81 on AIME2025), outperforming baselines with at least 2x fewer samples in 16 out of 20 comparisons. Our analysis reveals that correct reasoning chains exhibit significantly higher reasoning and answer confidence, justifying the effectiveness of PiCSAR.

Abstract

Feature Selection

Explore similar work

PiCSAR: Probabilistic Confidence Selection And Ranking for Reasoning Chains

When Do Biological Reasoning Models Use Their Biological Inputs?

ROOFS: RObust biOmarker Feature Selection