Biology subjects

Jaljuli, I.

Publications and source records attributed to Jaljuli, I..

2 recordsLinked to original sources

bioRxiv2025

Revisiting Tumor Mutational Burden Cutoff: Multi-Study Replicability in Immunotherapy

PURPOSETumor Mutational Burden (TMB) is a crucial biomarker for predicting the effectiveness of cancer immunotherapy. However, ongoing debates about the optimal cutoff significantly impact the clinical application of immunotherapy. The purpose of our study is to comprehensively evaluate TMB cutoffs for predicting immunotherapy outcomes across multiple studies, considering both cancer type and outcome endpoint, using statistically principled approaches. METHODSWe analyzed data from PredictIO, curated through a PubMed search for studies involving immune checkpoint blockade treatment in the adjuvant setting, excluding those involving combinations with chemotherapy, targeted treatment, or radiation. The included studies, published between January 2015 and June 2022, required tumor sequencing data from whole exomes or targeted gene panels for TMB assessment. Outcome endpoints included clinical benefit rate (CBR), progression-free survival (PFS), and overall survival (OS). CBR was defined as the rate of complete response, partial response, or stable disease lasting at least six months, according to RECIST 1.1 criteria. OS and PFS were defined as the interval from treatment initiation to death and disease progression, respectively. TMB was uniformly derived from tumor sequencing data, representing the count of nonsynonymous mutations relative to the target sequencing size. TMB cutoffs were evaluated for outcome associations using study replicability analysis, alongside individual-study analysis and random-effects meta-analysis for comparison. RESULTSThe data provided sufficient evidence of replicable outcome associations for specific TMB cutoffs in melanoma, lung cancer, and bladder cancer. The FDA-recommended cutoff of 10 mutations per megabase showed replicable associations in melanoma for OS (p-value < 0.01) and CBR (p-value < 0.01), though more replicable cutoffs were identified for the latter. Lower cutoffs of 4 and 2 were found to be replicable in lung cancer for CBR (p-value = 0.04) and in bladder cancer for OS (p-value < 0.01), respectively. No cutoff was deemed replicable for the other cancer type and outcome combinations, due to no association, inadequate power, or insufficient data. CONCLUSIONA pan-cancer cutoff of 10 mutations per megabase may not be optimal for predicting immunotherapy outcomes. Further studies are needed to determine appropriate cutoffs specific to cancer types and outcomes through statistically principled replicability analyses.

cancer biology↗

bioRxiv2021

Improving replicability using interaction with laboratories: a multi-lab experimental assessment

Experimentation with mouse and rat models has become a central strategy for discovering mammalian gene function, and for preclinical testing of pharmacological treatments, yet the utility of any findings critically depends on their replicability in other laboratories. In previous publications we proposed a statistical approach for estimating the inter-laboratory replicability of novel discoveries made in a single laboratory. We demonstrated that previous phenotyping results from multi-lab databases can be used to derive a Genotype-by-Lab (GxL) adjustment factor to greatly enhance the replicability of the single-lab findings, for similarly measured phenotypes, even before making the effort of replicating these finding in additional laboratories. This demonstration, however, still raised several important questions that could only be answered by an additional large-scale prospective experiment: 1) Does GxL-adjustment work in single-lab experiments that were not intended to be standardized across laboratories, and with genotypes that were not included in the previous experiments? And 2) Can it be used to adjust the results of pharmacological experiments? We investigated these questions by attempting to replicate, across three laboratories, results from five single-lab studies in the Mouse Phenome Database (MPD), offering 212 comparisons, including 60 involving a pharmacological treatment: 18 mg/kg/day fluoxetine. In addition, we define and use a dimensionless GxL factor, by dividing the GxL variance by the standard deviation between animals within groups, as a more robust vehicle to transfer the adjustment from the multi-lab analysis to very different labs and genotypes. For genotype comparisons, GxL-adjustment reduced the rate of non-replicable discoveries from 60% to 12%, for the price of reducing the power to make replicable discoveries from 87% to 66%. In absolute numbers, the adjustment prevented 23 non-replicable discoveries for the price of missing only three replicated ones. Tools and data needed for deployment of this method across other mouse experiments are publicly available in MPD. Our results further point at some phenotypes as more prone to produce non-replicable results, while others, known to be more difficult to measure, are as likely to produce replicable results (once adjusted) such as the physiological measure, body weight.

animal behavior and cognition↗