Week 7 HW: Genetic Circuits Part II
Assignment Part 1: Intracellular Artificial Neural Networks (IANNs)
What advantages do IANNs have over traditional genetic circuits, whose input/output behaviors are Boolean functions?
Traditional genetic circuits 😐 Traditional genetic circuits typically implement logic like AND, OR, NOT — meaning outputs are binary (on/off).
😐 Boolean circuits are limited to combinations of discrete logic rules.
😐 Traditional genetic circuits are hard-coded.
😐 Biological systems are inherently noisy (stochastic gene expression). Boolean circuits can fail if signals fluctuate around thresholds
IANNs 😊 Neural networks operate with continuous values, not just 0 or 1. This allows cells to respond proportionally to input concentrations and encode gradients and subtle differences in signals.
😊 IANNs can map complex environmental signals and multi-factor biological states.
😊 IANNs can be trained and adapted to new conditions.
😊 Neural networks distribute computation across many nodes and use weighted sums → more tolerant to noise.
Describe a useful application for an IANN; include a detailed description of input/output behavior, as well as any limitations an IANN might face to achieve your goal.
A useful application would be a smart cancer-detecting and responding therapeutic cell. It´s a living classifier that decides whether to trigger a treatment based on a complex molecular signature. The goal is to engineer a cell that detects whether it is in a tumor microenvironment and it activates a therapeutic response only when high confidence is reached.
Input (each input corresponds to a measurable molecular feature):
➡️ Surface protein markers (e.g. high HER2, EGFR)
➡️ Metabolic signals (high lactate → tumor glycolisis)
➡️ Oxigen level (low)
➡️ Inflammatory cytokines
➡️ Low pH level (acidic microenvironment)
IANN processes inputs using weighted gene regulation and combinatorial control. Biologically, each “neuron” is a gene whose expression depends on a weighted sum of inputs. Activation functions are implemented via cooperative binding and thresholding via repression/activation dynamics. Therefore, the network learns a nonlinear decision boundary and this allows detection of patterns, like:
“Moderate hypoxia + high lactate + mild inflammation = tumor”,
even if no single signal is decisive.
Output:
➡️ Expression of a therapeutic protein (e.g., cytokine, toxin, checkpoint inhibitor)
Limitations:
➡️ Network outputs may drift or become inconsistent
➡️ Difficult to “train” the network accurately because in electronics, weights are precise numbers. In cells, weights are promoter strength, binding affinity, degradation rates and these are hard to tune precisely and they are sensitive to context
➡️ Large circuits consume energy and resources. This leads to slower growth and evolutionary pressure to disable the circuit
Draw a diagram for an intracellular multilayer perceptron where layer 1 outputs an endoribonuclease that regulates a fluorescent protein output in layer 2.
References
- https://d1wqtxts1xzle7.cloudfront.net/34415923/Eluyode_DT-libre.pdf?1407773797=&response-content-disposition=inline%3B+filename%3DScholars_Research_Library_Comparative_st.pdf&Expires=1774290637&Signature=ZCAud8G9WvQouddiClQgtlBZpLtToyVSkyu45AEt8SLRpZKPVEolnvW-p9s0SfUJMcu4mrZxZDlTnn93bUv34VL5Nz9etoQJX3uNFYJBo58Go6eqAyymB05X~qSoi7T8I1eJH9DvNaZgOLyIcfB724kloAsogijGkWcH5~FCUPkvPMYzXPh596yjNVFefl4GhilZi~APAooLZRiFBErfAr39sBRYsLKfUwoRSLNJ1i3nUbiMu0oEl77XOwneTsqR9tcbhSKG-RL9QtvdxQyE92JCsGd4G3ZAza2N7Ika1Izoc8H9fGUf1sYgYGf1U~zknoTutvSwfSv1VxdzAE4fpg__&Key-Pair-Id=APKAJLOHF5GGSLRBV4ZA