- Jupyter Notebook 42.7%
- JavaScript 38.4%
- HTML 10%
- CSS 8.9%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
|
|
||
| .gitignore | ||
| app.js | ||
| cd19_muts.ipynb | ||
| grantham_matrix.js | ||
| index.html | ||
| LICENSE | ||
| README.md | ||
| references.js | ||
| style.css | ||
In this repo, we share a small check on the prior we found on the literature for our paper "The costimulatory domain influences CD19 CAR-T cell resistance development in B-cell malignancies". Here we just want to check whether the rate was compatible with our model. We just check whether assuming that mutations will happen randomly as cells divide, how likely it is to find at least one inactivating mutation in our cell culture, according to a given mutation rate.
Please, first check out our notebook. Here we did the main exploration. In addition, we added a small AI-generated interactive tool translating the initial work in the notebook to allow some collaborators explore some parameters.
Notice
This was just intended as a sanity check. Please do not use the tool as a proof of anything, since there are tons of limitations for discovery research. Here we just wanted to see the influece of length and that no other main covariates could bias the probability of inactivating mutations appearing at X generations by orders of magnitude.
How to Run the interactive page
- Download the Repository: Clone or download the entire repository to your local machine.
- Open
index.html: Navigate to the downloaded folder and open theindex.htmlfile in any modern web browser
Methodology
The estimation process incorporates several key factors, implemented primarily within app.js:
-
Gene Sequence Input:
- Users can paste a raw coding sequence or use a FASTA format.
- An example (Human CD19 cDNA) is provided for quick testing.
- The number of exons is required for splice site estimations.
-
Gene Length:
- The length of the coding sequence (CDS) is a fundamental input (retrieved automatically from the sequence). The probability of a gene being hit by a random mutation is directly proportional to its length, making this a dominant factor in the overall inactivation probability. We assume you know the probability of a random mutation per base per cell division (e.g., we used 1e-8 as a reference, provided by Williams et al, Nature Communications, 2020, which is adjustable via a slider in the user interface.
-
Single Nucleotide Variants (SNVs):
- The tool systematically considers all possible single nucleotide changes across the provided gene sequence.
- Each potential SNV is translated into its corresponding amino acid change (if any) using a standard genetic code table.
- Mutations are classified as:
- Synonymous: No change in amino acid.
- Missense: Results in a different amino acid.
- Nonsense: Introduces a premature stop codon.
- Start Codon Loss: Alters the initial ATG start codon.
- Stop Codon Loss: Eliminates a stop codon, leading to read-through.
-
Missense Mutation Impact (Grantham Score):
- The functional impact of missense mutations is assessed using the Grantham score. This score quantifies the physicochemical difference between the original and substituted amino acids.
- Three categories of missense severity (Conservative, Moderate, Severe) are defined based on Grantham distance thresholds. These thresholds are adjustable via sliders in the user interface.
- The Grantham score is a historically used proxy for missense impact. While more modern, computationally intensive tools like PolyPhen-2, SIFT, or CADD can offer higher accuracy in predicting the functional consequences of missense mutations, they are beyond the scope of this tool, which aims for rapid prior estimation. For this purpose, where gene length is a primary determinant, the Grantham score provides a computationally efficient and adequate approximation.
-
Splice Site Mutations:
- Mutations affecting canonical splice sites are known to frequently disrupt splicing and lead to gene inactivation.
- The tool considers the two bases at each end of an exon (i.e., four bases per splice junction: the last two bases of the exon and the first two bases of the intron for a donor site, and the last two bases of the intron and first two bases of the exon for an acceptor site). For a cDNA sequence, it estimates this based on the number of exons, effectively considering (Exon Count - 1) * 4 critical bases at splice junctions.
- The probability that a mutation at these sites is inactivating is user-adjustable via a slider (
spliceInactivationRate).
-
Indel Mutations:
- The tool considers both frameshift and in-frame indels.
- Frameshift Indels: Insertions or deletions not a multiple of three bases typically cause a frameshift, leading to a completely different and usually non-functional protein product. These are assumed to be highly inactivating (user-adjustable probability, defaults to high).
- In-frame Indels: Insertions or deletions of a multiple of three bases. Their impact is considered less severe than frameshifts but can still be damaging. The rate at which in-frame indels are damaging is also user-adjustable.
- The relative proportion of SNVs to indels, and frameshift to in-frame indels, are user-configurable parameters.
Scope and Limitations
- Approximation for Prior Estimation: This tool is designed to provide a reasonable prior estimate for gene inactivation. The simplifications made (e.g., Grantham score for missense, uniform mutation rates, estimated splice site impact) are justified by this scope. The impact of inactivation for prior estimation is often more dependent on gene length than on the precise functional prediction of every variant.
- Uniform Mutation Rate: The model assumes a uniform mutation rate across the gene, ignoring potential mutational hotspots, coldspots, or the influence of local genomic context (e.g., chromatin structure, replication timing).
- cDNA-Based Analysis: The analysis is primarily based on coding sequences. Splice site analysis is therefore an estimation based on exon count and user-defined inactivation rates rather than a direct analysis of genomic splice junctions and their specific sequences.
- No Complex Interactions
- Binary Inactivation
- Indels. No actual model on it, just the user decides the effect they can have.
Despite these limitations, the tool serves its intended purpose of providing a quick, interactive, and adjustable to check the impact of different covariates and how basically the gene length is what driver the phenomenon we are studying.
Citation
If your work relies on this repository or any part of our work, please cite our paper: The costimulatory domain influences CD19 CAR-T cell resistance development in B-cell malignancies
License
This project is licensed under the CC BY-NC 4.0 License. See the LICENSE file for details.