Biology subjects

Millikin, R. J.

Publications and source records attributed to Millikin, R. J..

4 recordsLinked to original sources

bioRxiv2026

Quantifying Evidence for Competing Biomedical Hypotheses using Large Language Models and Bayesian Analysis

Science fundamentally depends on the generation and testing of hypotheses, many of them controversial. An explosion in scientific literature has made evaluating hypotheses even within a domain a problem of scale, and risks slowing an already extensive consensus-building process. While this challenge has prompted interest in automated hypothesis evaluation tools, existing methods have not yet proven effective for comparing hypotheses. Here, we introduce KM-GPT-DCH, an algorithm that combines co-occurrence methods with large language models (LLMs) to develop a transparent and reproducible literature-based algorithm to compare controversial hypotheses using a structured scoring approach with Bayesian methods to estimate confidence. When testing the algorithm on historical controversial hypotheses previously decided, KM-GPT-DCH chooses the correct hypothesis with high confidence several years before the scientific community or public do so. We further apply the algorithm to compare twenty unresolved controversial hypothesis pairs providing guidance for future research. The method can help researchers and the public to evaluate biomedical hypotheses such as "Is it more likely that monoamine deficiency or inflammation causes depression?" It can also be used to assess and visualize historical trends in the scientific literature. A web-based implementation of the algorithm is freely available at https://skim.morgridge.org.

bioinformatics↗

bioRxiv2025

SKiM-GPT: Combining Biomedical Literature-Based Discovery with Large Language Model Hypothesis Evaluation

BackgroundGenerating and testing hypotheses is a critical aspect of biomedical science. Typically, researchers generate hypotheses by carefully analyzing available information and making logical connections, which are then tested. The accelerating growth of biomedical literature makes it increasingly difficult to keep pace with connections between biological entities emerging across biomedical research. Recently developed automated means of generating hypotheses can generate many more hypotheses than can be easily tested. One such approach involves literature based discovery (LBD) systems such as Serial KinderMiner (SKiM), which surfaces putative A B C links derived from term co occurrence. However, LBD systems leave three critical gaps: (i) they find statistical associations, not biological relationships; (ii) they can produce false positive leads; and (iii) they do not assess agreement with a hypothesis in question. As a result, LBD search results often require costly manual curation to be of practical utility to the researcher. Large language models (LLMs) have the potential to automate much of this curation step, but standalone LLMs are hampered by hallucinations, lack of transparency in information sources, and the inability to reference data not included in the training corpus. ResultsWe introduce SKiM-GPT, a retrieval-augmented generation (RAG) system that combines SKiMs co-occurrence search and retrieval with frontier LLMs to evaluate user- defined hypotheses. For every chosen A-B-C SKiM hit, SKiM-GPT retrieves appropriate PubMed abstract texts, filters out irrelevant abstracts with a fine-tuned relevance model, and prompts an LLM to evaluate the users hypothesis, given the relevant abstracts. Importantly, the SKiM-GPT system is transparent and human-verifiable: it displays the retrieved abstracts, the hypothesis score, and a justification for the score grounded in the texts and written in natural language. On a benchmark consisting of 14 disease-gene-drug hypotheses, SKiM-GPT achieves strong ordinal agreement with four expert biologists (Cohens {kappa} = 0.84), demonstrating its ability to replicate expert judgment. ConclusionsSKiM-GPT is open-source (https://github.com/stewart-lab/skimgpt) and available through a web interface (https://skim.morgridge.org), enabling both wet-lab and computational researchers to systematically and efficiently evaluate biomedical hypotheses at scale.

bioinformatics↗

bioRxiv2020

Spritz: A Proteogenomic Database Engine

Proteoforms are the workhorses of the cell, and subtle differences between their amino acid sequence or post-translational modifications (PTMs) can change their biological function. To most effectively identify and quantify proteoforms in genetically diverse samples by mass spectrometry (MS), it is advantageous to search the MS data against a sample-specific protein database that is tailored to the sample being analyzed, in that it contains the correct amino acid sequences and relevant PTMs for that sample. To this end, we have developed Spritz (https://smith-chem-wisc.github.io/Spritz/), an open-source software tool for generating protein databases annotated with sequence variations and PTMs. We provide a simple graphical user interface (GUI) for Windows and scripts that can be run on any operating system. Spritz automatically sets up and executes approximately 20 tools, which enable construction of a proteogenomic database from only raw RNA sequencing data. Sequence variations that are discovered in RNA sequencing data upon comparison to the Ensembl reference genome are annotated on proteins in these databases, and PTM annotations are transferred from UniProt. Modifications can also be discovered and added to the database using bottom-up mass spectrometry data and global PTM discovery in MetaMorpheus. We demonstrate that such sample-specific databases allow the identification of variant peptides, modified variant peptides, and variant proteoforms by searching bottom-up and top-down proteomic data from the Jurkat human T lymphocyte cell line and demonstrate the identification of phosphorylated variant sites with phosphoproteomic data from the U2OS human osteosarcoma cell line.

bioinformatics↗

bioRxiv2020

Precursor intensity-based label-free quantification software tools for proteomic and multi-omic analysis within the Galaxy Platform.

For mass spectrometry-based peptide and protein quantification, label-free quantification (LFQ) based on precursor mass peak (MS1) intensities is considered reliable due to its dynamic range, reproducibility, and accuracy. In LFQ workflows, protein abundance changes are inferred from peptide-level information, including microbial peptides (for metaproteomics) and peptides carrying post-translational modifications (for proteomics) and/or variant sequences (for proteogenomics). Multi-omics studies (such as proteogenomics and metaproteomics) rely on peptide detection and quantification to identify and quantify peptides that map to unique proteoforms and metaproteins. The Galaxy for proteomics (Galaxy-P) platform has proven useful for the development of accessible workflows to identify proteins in these complex multi-omic studies. However, proteomics workflows within the Galaxy platform have lacked well-tested label-free quantification tools. In this study, our main goals were to evaluate two recently published open-source LFQ tools and to implement them within the Galaxy platform, enabling their easy integration with established workflows. These two tools, moFF and FlashLFQ, were selected based on their described peptide quantification capabilities and amenability to Galaxy implementation. Through rigorous testing and communication with the tools developers, we gained insights into the software features necessary for maximizing the performance of each tool. Software features evaluated included: a) match-between-runs (MBR); b) using both Thermo .raw and HUPO standards .mzML file formats as input for improved quantification; c) use of containers and/or conda packages; d) parameters needed for analyzing large input datasets; and e) optimization and validation of software performance. This work 1) establishes a process for software implementation, optimization and validation within Galaxy; and 2) makes powerful new tools for LFQ available which should prove highly useful for a variety of proteomics and multi-omics applications employing the Galaxy platform.

bioinformatics↗