https://pages.htgaa.org/2026a/nour-abdelrahman/homework/week-07-hw-genetic-circuits-part-ii/index.htmlPart 1: Intracellular Artificial Neural Networks (IANNs) What advantages do IANNs have over traditional genetic circuits, whose input/output behaviors are Boolean functions? The main advantage IANNs hold over traditional genetic circuits is scalability and the ability to support multilayer networks for complex decision-making. Traditional genetic circuits limitations include poor predictability and the struggle to reliably program multiple functions simultaneously due to inherent scalability limitations. On the other hand, ANNs have good predictability offering improved robustness for complex designs. Because of multiple layers and non-linear activations, neural networks can model complex, non-linear decision boundaries Traditional genetic circuits have input/output behaviors that function as Boolean operations. They process discrete signals (ON/OFF, high/low expression) through logic gates like AND, OR, and NOT, producing binary outputs based on truth tables. Moreover, the output layer in the ANNs producing the final prediction may be binary, multi-class or a continuous value. 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. Application of CNNs: tumor and MSI detection in gastrointestinal cancer Convolutional Neural Networks (CNNs) are deep learning models designed to analyze structured grid-like data such as images. the CNNs were used as automatic tumor detector to predict MSI (Microsatellite instability) that determines if the patient with gastrointestinal cancer will respond will to immunotherapy. The authors used hematoxylin and eosin (H&E)-stained histology slides as an input For tumor detection in gastrointestinal cancer, the authors trained a convolutional neural network with deep residual learning (resnet18)12 model to classify tumor versus normal tissue by transfer learning. Transfer learning means reusing a pre-trained neural network model on a new but related task, instead of training from scratch. For MSI detection, we trained another resnet18 model for each tumor type. input/output behavior Input: Tiles extracted from digitized histology slides. Output: For each tile, a probability score indicating tumor vs. normal or MSI vs. MSS status. Behavior: The neural network processes image features within each tile to generate these probability scores, enabling localized tissue characterization and subsequent patient-level molecular classification. The mentioned limitations of CNN were: Classifying ability is limited to cancer type and ethnicity in the training set. therefore, larger training cohorts are needed to boost classification performance because rare morphological variants can be learned by the network The required tissue size. To define its lower limit, they generated ‘virtual biopsies’ and found that performance plateaued at approximately 100 tiles of 256 μm edge length, suggesting that biopsies are sufficient for MSI prediction Below is a diagram depicting an intracellular single-layer perceptron where the X1 input is DNA encoding for the Csy4 endoribonuclease and the X2 input is DNA encoding for a fluorescent protein output whose mRNA is regulated by Csy4. Tx: transcription; Tl: translation. ReferencesHugoen