# colorectal cancer

Whether colorectal cancer recurs after resection has been judged mainly from routine pathology such as lymphovascular invasion, perineural invasion, tumor budding and Ki-67. AlphaFISH pairs a sequencing-free nanoprobe biochip with a transformer model to read 41 host transcripts and 51 bacterial genera across 149 colorectal biopsies, and in 28 patients followed for one year after surgery it predicted recurrence or metastasis with an AUC of 0.81, against 0.45 for the combined routine clinical markers. The result comes from a preprint that has not been peer reviewed, on a limited sample.

Whether colorectal cancer recurs after resection has been judged mainly from routine pathology such as lymphovascular invasion, perineural invasion, tumor budding and Ki-67. AlphaFISH pairs a sequencing-free nanoprobe biochip with a transformer model to read 41 host transcripts and 51 bacterial genera across 149 colorectal biopsies, and in 28 patients followed for one year after surgery it predicted recurrence or metastasis with an AUC of 0.81, against 0.45 for the combined routine clinical markers. The result comes from a preprint that has not been peer reviewed, on a limited sample.

Using 63 organoid samples from 22 colorectal cancer patients, external validation in TCGA-COAD/READ (n=624) and GSE39582 (n=536) totaling 1,160 cases, public cell line panels (GDSC2, DepMap), and 65 lines from an independent patient-derived CRC organoid biobank, the study tested whether usage of the dual MDM2 promoters (P1/P2) acts as a molecular switch separating a chromosomal-instability type from an environment-adaptive type (microsatellite instability/serrated pathway with gastric metaplasia), finding deep-learning morphological classification at 98.5% test accuracy (64/65), morphology corresponding to P1/P2 isoform usage (median Type1 fraction 0.826 versus 0.444 in P1-dominant samples; non-Type1 cystic mucinous morphology in P2-dominant samples, AUC 0.

Using 63 organoid samples from 22 colorectal cancer patients, external validation in TCGA-COAD/READ (n=624) and GSE39582 (n=536) totaling 1,160 cases, public cell line panels (GDSC2, DepMap), and 65 lines from an independent patient-derived CRC organoid biobank, the study tested whether usage of the dual MDM2 promoters (P1/P2) acts as a molecular switch separating a chromosomal-instability type from an environment-adaptive type (microsatellite instability/serrated pathway with gastric metaplasia), finding deep-learning morphological classification at 98.5% test accuracy (64/65), morphology corresponding to P1/P2 isoform usage (median Type1 fraction 0.826 versus 0.444 in P1-dominant samples; non-Type1 cystic mucinous morphology in P2-dominant samples, AUC 0.

Using 2,704 colorectal cancer discharge notes from MIMIC-IV and a 46-symptom inventory derived from the MSAS and EORTC QLQ-CR29, this study benchmarked dictionary-based rule matching, pretrained clinical NER, zero-shot Claude Haiku and Gemini 3.5 Flash, and two hybrid variants (LLM output plus post-hoc rule-based negation filtering) against a 200-note two-rater adjudicated gold standard, finding that Gemini 3.5 Flash performed best (Macro F1=0.70, Micro F1=0.86, Macro Precision=0.74), followed by Claude Haiku (Macro F1=0.63, Macro Recall=0.71), both substantially outperforming rule-based (Macro F1=0.44) and NER (Macro F1=0.38) methods, while post-hoc negation filtering paradoxically degraded LLM performance (Gemini+Hybrid Macro F1=0.58; Claude+Hybrid Macro F1=0.54).

Using 2,704 colorectal cancer discharge notes from MIMIC-IV and a 46-symptom inventory derived from the MSAS and EORTC QLQ-CR29, this study benchmarked dictionary-based rule matching, pretrained clinical NER, zero-shot Claude Haiku and Gemini 3.5 Flash, and two hybrid variants (LLM output plus post-hoc rule-based negation filtering) against a 200-note two-rater adjudicated gold standard, finding that Gemini 3.5 Flash performed best (Macro F1=0.70, Micro F1=0.86, Macro Precision=0.74), followed by Claude Haiku (Macro F1=0.63, Macro Recall=0.71), both substantially outperforming rule-based (Macro F1=0.44) and NER (Macro F1=0.38) methods, while post-hoc negation filtering paradoxically degraded LLM performance (Gemini+Hybrid Macro F1=0.58; Claude+Hybrid Macro F1=0.54).