Compare neutral drift model vs frequency-dependent selection for ARG frequency distributions in 3 databases (CARD, ResFinder, AMRFinderPlus) across 2,400 bacterial genomes. Neutral drift (Wright-Fisher with mutation) fits observed frequency spectra with KS p>0.
Compare CLR, ALR, ILR, and raw relative abundance on 4 published microbiome-disease association datasets (IBD, obesity, colorectal cancer, diabetes). The 'winning' method (highest number of significant associations at FDR<0.
Benchmark ML survival models (Cox-PH, RSF, DeepSurv, Cox-nnet) on genomics/transcriptomics/proteomics features vs TNM clinical staging alone across 12 TCGA cohorts (N=5,847). Mean C-index: clinical staging 0.
Apply rigorous statistical tests (Clauset-Shalizi-Newman framework) to degree distributions of 6 PPI databases (BioGRID, STRING, IntAct, MINT, DIP, HPRD). Power-law fits are rejected (p<0.
Compare Cellpose, StarDist, Mesmer, DeepCell, ACDC on 10 tissue types (H&E, 2000 images). Overall Dice agreement 0.
Compare GRCh38, T2T-CHM13, and HPRC pangenome reference on GTEx v8 data (838 samples, 49 tissues). 15.
Compare 5 SV callers (Manta, Delly, GRIDSS, Sniffles, cuteSV) on GIAB HG002 truth set. Insertions >1kb: pairwise concordance 38-47%.
Batch effects are a major confounder in genomics, and multiple correction methods exist. We compare ComBat, limma removeBatchEffect, Harmony, scVI, and MNN on 5 paired RNA-seq datasets where the same biological comparison was performed in two independent batches.
Alternative polyadenylation (APA) has been proposed as a cancer biomarker, with studies reporting widespread 3'UTR shortening in tumors. We test whether APA changes are cancer-specific or tissue-specific by analyzing RNA-seq data from 8 TCGA cancer types across 5 tissue origins (4,200 tumor, 800 normal samples).
GC-content bias in microarray and RNA-seq platforms is well-documented but rarely corrected in differential expression analyses. We audit 20 widely-cited microarray datasets from GEO, applying a permutation-based test that evaluates whether the overlap between differentially expressed gene lists and GC-content-correlated genes exceeds chance.
SNNs promise energy efficiency via sparse spike trains, but accuracy requires sufficient timesteps, creating a latency-accuracy tradeoff. We characterize this for 8 SNN architectures on CIFAR-10/100 and DVS-Gesture at timesteps 1-128.
The sim-to-real transfer gap is assumed to grow with task complexity, but we find a U-shaped relationship. Across 6 manipulation tasks (reaching, pushing, pick-and-place, stacking, insertion, bimanual assembly) with 5 domain randomization levels on Franka Emika: simple tasks transfer well (gap 8-12%), moderate tasks show maximum gap (28-41%), complex tasks show reduced gap (18-24%).
In cooperative MARL, free-riding agents contribute minimally while benefiting from team rewards. We propose Shapley Contribution Tracking (SCT) using online Shapley value approximation.
Multi-agent LLM systems chain multiple model instances via natural language, but scaling properties are unknown. We study 2-16 agents across four patterns (sequential, broadcast, hierarchical, peer-to-peer).
Fault-tolerant LLM training requires periodic checkpointing. We analyze the cost structure across 64-4,096 GPUs, comparing checkpoint overhead against failure recovery cost.
Distributed LLM training suffers from straggler nodes that impose synchronization barriers. We analyze 2,400 training runs on clusters of 10-512 GPUs with data/tensor/pipeline parallelism.
LLM APIs process inputs autoregressively, coupling response latency to input/output length. We demonstrate this creates an exploitable timing side channel: observing only response time reveals input token count with 93.
Prompt injection is a critical LLM security vulnerability. We analyze the tradeoff between injection resistance and helpfulness across 12 models from 4 families.
Technical debt density distribution across projects is poorly understood. We analyze 8,247 projects (6 languages) via SonarQube.
LLMs generate unit tests with impressive coverage, but we challenge this optimism using mutation testing. We evaluate GPT-4, Claude-3, CodeLlama-34B, and DeepSeek-Coder-33B on 200 Python functions from popular libraries.